[go: up one dir, main page]

US20070163425A1 - Melody retrieval system - Google Patents

Melody retrieval system Download PDF

Info

Publication number
US20070163425A1
US20070163425A1 US09/804,069 US80406901A US2007163425A1 US 20070163425 A1 US20070163425 A1 US 20070163425A1 US 80406901 A US80406901 A US 80406901A US 2007163425 A1 US2007163425 A1 US 2007163425A1
Authority
US
United States
Prior art keywords
melody
pitch
note
frame
breakpoints
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Abandoned
Application number
US09/804,069
Other languages
English (en)
Inventor
Chi-ying Tsui
Bertram Shi
Chi Yung
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Perception Digital Technology BVI Ltd
Original Assignee
Perception Digital Technology BVI Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Perception Digital Technology BVI Ltd filed Critical Perception Digital Technology BVI Ltd
Priority to US09/804,069 priority Critical patent/US20070163425A1/en
Assigned to PERCEPTION DIGITAL TECHNOLOGY (BVI) LIMITED reassignment PERCEPTION DIGITAL TECHNOLOGY (BVI) LIMITED ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: YUNG, CHI WAI, SHI, BERTRAM, TSUI, CHI-YING
Publication of US20070163425A1 publication Critical patent/US20070163425A1/en
Priority to US11/953,215 priority patent/US7919706B2/en
Abandoned legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H3/00—Instruments in which the tones are generated by electromechanical means
    • G10H3/12—Instruments in which the tones are generated by electromechanical means using mechanical resonant generators, e.g. strings or percussive instruments, the tones of which are picked up by electromechanical transducers, the electrical signals being further manipulated or amplified and subsequently converted to sound by a loudspeaker or equivalent instrument
    • G10H3/125—Extracting or recognising the pitch or fundamental frequency of the picked up signal
    • G—PHYSICS
    • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
    • G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
    • G09B15/00—Teaching music
    • G09B15/02—Boards or like means for providing an indication of notes
    • G09B15/023—Electrically operated
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H1/00—Details of electrophonic musical instruments
    • G10H1/0008—Associated control or indicating means
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2210/00—Aspects or methods of musical processing having intrinsic musical character, i.e. involving musical theory or musical parameters or relying on musical knowledge, as applied in electrophonic musical tools or instruments
    • G10H2210/031—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal
    • G10H2210/066—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal for pitch analysis as part of wider processing for musical purposes, e.g. transcription, musical performance evaluation; Pitch recognition, e.g. in polyphonic sounds; Estimation or use of missing fundamental
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2210/00—Aspects or methods of musical processing having intrinsic musical character, i.e. involving musical theory or musical parameters or relying on musical knowledge, as applied in electrophonic musical tools or instruments
    • G10H2210/031—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal
    • G10H2210/076—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal for extraction of timing, tempo; Beat detection
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2240/00—Data organisation or data communication aspects, specifically adapted for electrophonic musical tools or instruments
    • G10H2240/095—Identification code, e.g. ISWC for musical works; Identification dataset
    • G10H2240/101—User identification
    • G10H2240/111—User Password, i.e. security arrangements to prevent third party unauthorised use, e.g. password, id number, code, pin
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2240/00—Data organisation or data communication aspects, specifically adapted for electrophonic musical tools or instruments
    • G10H2240/121—Musical libraries, i.e. musical databases indexed by musical parameters, wavetables, indexing schemes using musical parameters, musical rule bases or knowledge bases, e.g. for automatic composing methods
    • G10H2240/131—Library retrieval, i.e. searching a database or selecting a specific musical piece, segment, pattern, rule or parameter set
    • G10H2240/141—Library retrieval matching, i.e. any of the steps of matching an inputted segment or phrase with musical database contents, e.g. query by humming, singing or playing; the steps may include, e.g. musical analysis of the input, musical feature extraction, query formulation, or details of the retrieval process
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2240/00—Data organisation or data communication aspects, specifically adapted for electrophonic musical tools or instruments
    • G10H2240/171—Transmission of musical instrument data, control or status information; Transmission, remote access or control of music data for electrophonic musical instruments
    • G10H2240/201—Physical layer or hardware aspects of transmission to or from an electrophonic musical instrument, e.g. voltage levels, bit streams, code words or symbols over a physical link connecting network nodes or instruments
    • G10H2240/241—Telephone transmission, i.e. using twisted pair telephone lines or any type of telephone network
    • G10H2240/251—Mobile telephone transmission, i.e. transmitting, accessing or controlling music data wirelessly via a wireless or mobile telephone receiver, analogue or digital, e.g. DECT, GSM, UMTS
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2240/00—Data organisation or data communication aspects, specifically adapted for electrophonic musical tools or instruments
    • G10H2240/171—Transmission of musical instrument data, control or status information; Transmission, remote access or control of music data for electrophonic musical instruments
    • G10H2240/281—Protocol or standard connector for transmission of analog or digital data to or from an electrophonic musical instrument
    • G10H2240/295—Packet switched network, e.g. token ring
    • G10H2240/305—Internet or TCP/IP protocol use for any electrophonic musical instrument data or musical parameter transmission purposes
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2250/00—Aspects of algorithms or signal processing methods without intrinsic musical character, yet specifically adapted for or used in electrophonic musical processing
    • G10H2250/025—Envelope processing of music signals in, e.g. time domain, transform domain or cepstrum domain
    • G10H2250/031—Spectrum envelope processing
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2250/00—Aspects of algorithms or signal processing methods without intrinsic musical character, yet specifically adapted for or used in electrophonic musical processing
    • G10H2250/131—Mathematical functions for musical analysis, processing, synthesis or composition
    • G10H2250/135—Autocorrelation
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
    • G10H2250/00—Aspects of algorithms or signal processing methods without intrinsic musical character, yet specifically adapted for or used in electrophonic musical processing
    • G10H2250/541—Details of musical waveform synthesis, i.e. audio waveshape processing from individual wavetable samples, independently of their origin or of the sound they represent
    • G10H2250/645—Waveform scaling, i.e. amplitude value normalisation

Definitions

  • the invention relates to the field of music retrieval systems and more particularly to retrieval systems which take a melody vocalized by a user as the query.
  • the invention provides an approach different from those described in the above-mentioned documents in identifying a musical composition in response to a query that is a melody.
  • the invention provides methods and systems for retrieving musical selections or data identifying musical selections based on a digital version of a melody which originated from a sound or electronic source, e.g., a person humming, singing, whistling or otherwise vocalizing the melody; a musical instrument's audio or electronic output; an analog or digital recording of the melody, etc. Breakpoints between notes are identified as are distinct notes represented by pitch. In addition, one or more confidence levels may be associated with the input melody.
  • a value or confidence level may be assigned to each breakpoint to provide a measure of confidence that the identified breakpoint is in fact a breakpoint.
  • a value or confidence level assigned to each note may provide a measure of confidence that the identified note is a single note, e.g., does not include two or more notes.
  • One aspect of the invention provides a method and system for converting a digitized melody into a series of notes.
  • the method and system receive a digitized representation of an input melody, identify breakpoints in the melody in order to define notes therein, determine a pitch and beat duration for each note of the melody, and associate a confidence level with each breakpoint, or each note, or both.
  • the confidence levels associated with breakpoints and/or notes may be determined using different techniques, some of which are described herein.
  • segmentation of the input melody into distinct notes divided by breakpoints is based on changes or differences in the distribution of energy across the frequency spectrum over time.
  • the confidence levels associated with each breakpoint and/or note may be based on changes in pitch, as well as absolute and relative values of a spectral energy distribution indicator.
  • One aspect of the invention provides a method and related system for converting a digitized melody into a sequence of notes.
  • the method involves estimating breakpoints in the input melody based on changes in the distribution of energy across the frequency spectrum over time.
  • the melody is segmented into a series of frames.
  • a spectral energy distribution (SED) indicator is computed for each frame and at least initial breakpoints estimates are derived based on the SED indicator. Notes are defined between adjacent breakpoints.
  • Another aspect of the invention provides another method and related system for converting a digitized melody into a sequence of notes.
  • the method includes: segmenting the melody into a series of frames; computing the auto-correlation of each frame; estimating the pitch of each frame based on (i) a pitch period corresponding to a shift where the auto-correlation coefficient associated with the frame is relatively large and (ii) the closeness of the pitch estimate to estimates in one or more adjacent frames; and estimating breakpoints in the melody based on changes in the pitch estimates, wherein the notes are defined between adjacent breakpoints.
  • Another aspect of the invention provides a method and related system for identifying breakpoints in a digitized melody.
  • the method includes: segmenting the melody into a series of frames; computing the auto-correlation of each frame; estimating the pitch of each frame based on (i) a pitch period corresponding to a shift where the auto-correlation coefficient associated with the frame is relatively large and (ii) the closeness of the pitch estimate to estimates in one or more adjacent frames; determining regions of said melody where pitch estimates are likely to be invalid; and identifying the breakpoints in the melody based on transitions between frames having valid pitch estimates and transitions having invalid pitch estimates.
  • Other aspects of the invention relate to methods and systems for determining confidence levels for breakpoints and/or notes in a waveform representing a melody. These methods include segmenting the waveform into a series of frames, wherein adjacent breakpoints encompass one or more sequential frames, each note being defined between adjacent breakpoints. Then, at least one of the following three steps may be executed: (a) computing a spectral energy distribution (SED) indicator for each frame; (b) estimating the pitch of each frame; and (c) determining the energy level of each frame.
  • the confidence levels may be based on any of the following three characteristics: (i) the SED indicator, (ii) changes in pitch, and (iii) the energy level.
  • An entry may be retrieved from a music database of sequences of pitches and beat durations in accordance with a match function that receives the digitized melody obtained from a melody source as described herein.
  • a method and system for implementing the retrieval may determine a score for each entry based on a search which minimizes the cost of matching the pitches and beat durations of the melody and the entry, and which may be based on minimizing a cost computation which may take into account one or more note insertion and/or deletion errors and penalize the cost in accordance with confidence levels pertaining thereto.
  • Another aspect of the invention relates to a method and system of retrieving at least one entry from a music database, wherein each entry is associated with a sequence of pitches and beat durations.
  • the method includes receiving a digitized representation of an input melody; identifying breakpoints in the melody in order to define notes therein; associating each breakpoint and/or note with a confidence level; and determining a pitch and beat duration for each note of the melody. Then, a score is determined for each database entry based on a search which minimizes the cost of matching the pitches and beat durations of the melody and the entry.
  • the search considers at least one deletion or insertion error in a selected note of the melody and, in this event, penalizes the cost of matching based on the confidence level of the selected note or breakpoint associated therewith. At least one entry may then be presented to a user based on its score.
  • FIG. 1 is a system block diagram showing the major components of a music recognition system according to a preferred embodiment of the invention
  • FIG. 2 is a functional block diagram showing the processing blocks of a melody-to-note conversion subsystem employed in the music recognition system of FIG. 1 ;
  • FIG. 3 is a schematic diagram illustrating some of the processing activities of the melody-to-note conversion subsystem with respect to a sample input melody
  • FIG. 4A is a normalized energy spectrogram, plotted against time and frequency, of a sample input melody (which sample differs from the melody referenced in FIG. 3 );
  • FIG. 4B is a graph of the normalized energy spectrum at a first time frame in FIG. 4A plotted against frequency
  • FIG. 4C is a graph of the normalized energy spectrum at a second time frame in FIG. 4A plotted against frequency
  • FIG. 5A is identical to FIG. 4A (and provided on the same drawing sheet as FIGS. 5B and 5C for reference purposes);
  • FIG. 5B is a graph of a spectral energy distribution indicator, computed in a first manner, which is based upon the spectrogram of FIG. 5A ;
  • FIG. 5C is a graph of a “minimum measure”, as discussed in greater detail below, which is based on the spectral energy distribution indicator shown in FIG. 5B ;
  • FIG. 6A is identical to FIG. 4A (and provided on the same drawing sheet as FIGS. 6B and 6C for reference purposes);
  • FIG. 6B is a graph of a spectral energy distribution indicator, computed in a second manner, which is based upon the spectrogram of FIG. 6A ;
  • FIG. 6C is a graph of a “minimum measure”, as discussed in greater detail below, which is based on the spectral energy distribution indicator shown in FIG. 6B ;
  • FIG. 7A is identical to FIG. 4A (and provided on the same drawing sheet as FIGS. 7B and 7C for reference purposes);
  • FIG. 7B is a graph of a spectral energy distribution indicator, computed in a third manner, which is based upon the spectrogram of FIG. 7A ;
  • FIG. 7C is a graph of a “minimum measure”, as discussed in greater detail below, which is based on the spectral energy distribution indicator shown in FIG. 7B ;
  • FIG. 8A is identical to FIG. 4A (and provided on the same drawing sheet as FIGS. 8B and 8C for reference purposes);
  • FIG. 8B is a graph of a spectral energy distribution indicator, computed in a fourth manner, which is based upon the spectrogram of FIG. 8A ;
  • FIG. 8C is a graph of a “minimum measure”, as discussed in greater detail below, which is based on the spectral energy distribution indicator shown in FIG. 8B ;
  • FIG. 9A is identical to FIG. 4A (and provided on the same drawing sheet as FIGS. 9B and 9C for reference purposes);
  • FIG. 9B is a graph of a spectral energy distribution indicator, computed in a fifth manner, which is based upon the spectrogram of FIG. 9A ;
  • FIG. 9C is a graph of a “minimum measure”, as discussed in greater detail below, which is based on the spectral energy distribution indicator shown in FIG. 9B ;
  • FIG. 10 is a schematic diagram illustrating a process for matching notes.
  • FIG. 1 shows a music recognition system 10 which comprises four major components: a melody-to-note conversion subsystem 12 ; a music reference database 14 ; a note-matching engine 16 ; and an output subsystem 18 .
  • the music recognition system 10 takes a digitized input melody 20 obtained from a source 11 as a query.
  • the input melody originate from a user in the form of humming, particularly through intonations of notes that are combinations of a semi-vowel, such as “l”, and vowel, such as “a” (i.e., notes in the form of “la”).
  • the input melody may also comprise many other forms of humming, singing, whistling or other such types of music-like vocalization.
  • the input melody may also originate from a musical instrument(s). In these cases the source 11 represents circuitry for recording and digitizing the user's voice or the musical instrument.
  • the input melody may originate from a recording of some kind, in which case the source 11 represents the corresponding player and, if necessary, any circuitry for digitizing the output of the player.
  • the digitized input melody 20 is supplied to the melody-to-note conversion subsystem 12 .
  • the melody-to-note conversion subsystem 12 converts the digitized input melody 20 into a sequence of musical notes characterized by pitch, beat duration and confidence levels. This is accomplished through spectral analysis techniques described in greater detail below which are used to find “breakpoints” in the input melody in order to separate it into distinct notes.
  • the pitch of each note is determined by the periodicity of the input melody waveform between the note-defining breakpoints.
  • the beat duration of each note is extracted from the separation of the notes, i.e., the duration is determined from the time period between breakpoints.
  • each breakpoint is preferably associated with a confidence level, which indicates how likely the breakpoint is a valid breakpoint.
  • a confidence level is preferably also associated with each note to indicate how unlikely the identified note actually contains more than one note.
  • the output of the melody-to-note conversion subsystem 12 is a differential note and timing file 150 which comprises the relative difference in pitch and the relative difference in beat duration of consecutive notes.
  • the difference is preferably expressed in terms of the logarithm of the ratio of the pitch and duration values of the consecutive notes. The reason for using pitch and duration differences is discussed further below.
  • the music reference database 14 stores the differential note and timing files of all music or songs searchable by the system 10 .
  • Each such file preferably comprises a short, easily recognizable segment of a song or music, i.e., the so-called “signature melody”, but may alternatively encompass an entire song or piece of music.
  • These files may be generated in the first instance by the melody-to-note conversion subsystem 12 .
  • the note matching engine 16 compares the differential note and timing file 150 from the melody-to-note conversion subsystem 12 with songs or pieces of music in the music reference database 14 , which are stored in a similar file format. Since different users may vocalize or play a song or piece of music in different key and different tempo, the system 10 does not compare the pitch of the uttered melody and the reference files directly, but rather the ratio in pitch between consecutive notes. For the same melody, if the scale is shifted to a different frequency, the ratio in the frequency (pitch) of the consecutive notes will be the same. Similarly, to normalize for differences in tempo, the system 10 compares the relative duration of the consecutive notes.
  • the note matching engine 16 employs dynamic programming techniques described in greater detail below for matching the differential note and timing file 150 with similarly formatted files stored in the music database 14 . These techniques can compensate for pitch errors and insertions or deletions of notes by the user or the melody-to-note conversion subsystem 12 .
  • the engine 16 calculates a matching score for each song in the database 14 .
  • the output subsystem 18 sorts the songs or music in the database 16 based on the matching scores. The highest ranked song(s) or piece(es) of music is selected for presentation to the user.
  • FIG. 2 shows the functional blocks of the melody-to-note conversion subsystem 12 .
  • the subsystem 12 generates the following data from the digitized input melody 20 , which is used to construct the output differential note and timing file 150 :
  • the subsystem 12 determines one or more confidence levels related to breakpoints and/or notes, and uses one or more of those confidence levels in the construction of the differential note and timing files 150 .
  • a confidence measure or level is associated with each breakpoint that indicates the probability that the breakpoint is valid.
  • a confidence measure or level may also be associated with each identified note, which indicates the likelihood that the identified note does not contain more than one note.
  • Breakpoints are intended to indicate points of silence or points of inflection (i.e., alteration in pitch or tone of the voice) in the input melody.
  • the embodiments described herein use more than one technique to identify a breakpoint and determine its confidence level by considering how “closely” the various techniques have collectively identified a breakpoint. For example, if all techniques have identified a breakpoint at the same particular point in the input melody, the confidence level associated with that breakpoint is relatively high. Conversely, if one or less than all of the techniques do not identify a breakpoint at or near that particular point in the melody, the confidence level will be lower.
  • the first two characteristics should be intuitively understood for their value in identifying a breakpoint.
  • the last item is a breakpoint characteristic due to the typical nature of human vocalization. More particularly, as mentioned, users can hum melodies using notes which are combination of a semi-vowel, such as “l” and a vowel, such as “a”, i.e. “la.”. When enunciating the semi-vowel, it has been found that the mouth is typically actuated in such as way that results in the sound energy being concentrated at lower frequencies, as compared with the frequency distribution of the sound energy during the vowel. The preferred embodiment takes advantage of this observation, as discussed in greater detail below.
  • note confidence measures can be derived from the same quantities as used for breakpoint confidence measures, except with an inverse relationship. For example, a large and rapid change in pitch near a breakpoint increases the confidence in that breakpoint. However, large and rapid changes in pitch in the interval between two breakpoints decreases the confidence that a breakpoint has not been missed. As with breakpoint confidence measures, note confidence measures may be based on one or more different indicators.
  • One set of processing steps of the subsystem 12 begins by filtering the input melody 20 (alternatively referred to as the “input waveform”) with a bandpass filter 25 in order to attenuate frequency components that lie outside the range of expected pitches.
  • a framer 30 segments the filtered input waveform into a sequence of “frames” of equal period, e.g., 1/32 of a second. Each frame contains a short portion of the total filtered input waveform. Adjacent frames may contain overlapping parts of the filtered input waveform to provide for some degree of continuity therebetween, as known in the art per se. The overlap is preferably a tunable parameter and may be expressed as a percentage. Every part of the filtered input waveform is thus represented in at least one frame.
  • the auto-correlation of each frame is then computed at block 35 .
  • This provides a measure of the similarity of a signal with a shifted version of itself, where the amount of shift is given by l.
  • the auto-correlation is related to the spectral energy distribution of x[n].
  • the auto-correlation computation will yield a multitude of auto-correlation coefficients for each frame. As known in the art, peaks in the auto-correlation provide an indication of the periodicity or pitch of a waveform, which in this case is the part of the filtered input waveform contained in each frame.
  • Block 45 provides a frame-by-frame pitch estimate 50 . This is carried out by first identifying the “largest” peaks in the auto-correlation of each frame, e.g., the top 2-10 auto-correlation values. This yields a number of “pitch period candidates”.
  • the estimated pitch period of the frame is determined by selecting the pitch period candidate that corresponds to a large auto-correlation peak while simultaneously considering how “close” the pitch period candidate is to pitch period estimates in one or more adjacent frames.
  • the adjacent frames may be preceding or receding frames, or both.
  • the preferred embodiment employs a cost function which weights the size of the auto-correlation peaks and the closeness of the corresponding pitch period candidates to pitch period estimates in adjacent frames.
  • p i be the pitch period in frame i, where p i is either one of the identified pitch period candidates or a value indicating the lack of any identified pitch.
  • An example cost function is ⁇ i ⁇ ⁇ D ⁇ ( p i - 1 , p i ) + ( 1 - c ⁇ [ p i ] ) ⁇ where the sum is taken over all frames in the input melody.
  • where ⁇ 2/ln2.
  • D(p i ⁇ 1 , p i ) is set equal to a constant, e.g. 4.
  • the exact sequence of pitch period candidates minimizing this cost function can be computed by a dynamic programming procedure similar to that described in Section 3 on the note matching engine.
  • Block 55 seeks to detect regions 60 of the input waveform containing useful sound such as humming or music (as opposed to silence or noise), based on the frame-by-frame pitch estimate 50 and the frame-based auto-correlation 40 .
  • the manner in which this is preferably carried out is exemplified in FIG. 3 .
  • each position along the horizontal axis represents a frame, with the “P” line 56 representing input pitch estimates 50 ( FIG. 2 ) and the “E” line 57 representing the energy of each frame, as determined from the frame-based auto-correlation 40 ( FIG. 2 ).
  • the pitch estimates and energy estimates have quantized values ranging from 1-9.
  • the sound detection block 50 first looks for regions that may have useful sound because a valid pitch estimate was computed in the block 45 . This is shown in line “S 1 ” 58 of FIG. 3 where the symbol ‘H’ represents useful sound.
  • the sound detection block 55 considers the average energy of the frames in each region. Where the average energy is below a specified threshold, the region is considered to have no useful sound. This is shown in line “S 2 ” 59 of FIG. 3 .
  • the block 50 thus considers region 60 B of the input waveform as being silent. Conversely, regions 60 A and 60 C are considered to contain useful sound. Regions containing useful sound are sent to a breakpoint detection block 100 .
  • the breakpoint detection block 100 ( FIG. 2 ) also receives input from a parallel processing path comprising a high-pass filter 65 , a framer 70 and a spectral energy distribution indicator computation block 75 .
  • the high-pass filter 65 filters the input waveform 20 in order to emphasize high frequency information that has been found to be useful in detecting the breakpoints between notes.
  • the framer 70 slices the filtered input waveform into frames, which are identical in scope and temporal position to the frames generated by framer 30 .
  • the spectral energy distribution (“SED”) indicator computation block 75 computes a numerical measure or SED indicator 90 , which indicates how the sound energy is distributed in each frame.
  • the SED indicator preferably assumes relatively high values if the sound energy is concentrated near high frequencies and relatively low values if the sound energy is concentrated near other frequencies, as described in greater detail below. For example, a 4 kHz frequency range may be considered with high frequencies deemed to those approaching 4 kHz and low frequencies deemed to be those near zero kHz.
  • the breakpoint detection block 100 finds initial estimates for the locations of note breakpoints (i.e., “candidate” breakpoints) 105 and computes a confidence measure 110 associated with each candidate breakpoint 105 . This confidence measure varies between 0 and 1, where a value near 1 indicates that the breakpoint is very reliable.
  • the breakpoint detection block 100 operates on regions of the input waveform supplied by the sound detection block 55 . In FIG. 3 , for example, these would be regions 60 A and 60 C.
  • the detection block 100 assigns breakpoints to the beginning and end frames of each region. Thus, the transitions between a frame with no pitch estimate and a frame with a valid pitch estimate is one method that may be used to identify breakpoints. These breakpoints are given a confidence level of 1. This is exemplified in FIG. 3 by the “x” symbol in the “B” line 101 .
  • the block 100 detects candidate breakpoints based on minima present in the SED indicator 90 . These are exemplified in FIG. 3 by the “ ⁇ ” symbol in the “B” line 101 .
  • the reason for this can be understood on an intuitive level by considering a melody waveform that consists of a sequence of notes, each of which is sung as “la.”
  • the vowel part “a” is typically longer in duration than the consonant part and is usually better defined spectrally. Therefore, it should provide the most reliable information for pitch extraction. Segmentation can be performed if the “l” part of each “la” can be detected. Because “ 1 ” is a semivowel, it typically contains strong pitch periodicity. However, because of the constriction of the mouth during production, it contains less overall energy and less high frequency resonant structure.
  • FIG. 4A is a spectrogram of a normalized energy spectrum for a sample melody hummed using “la” notes.
  • the normalized energy spectrum is shown as a gray scale image wherein normalized energy values approaching a maximum value are white and values near zero are black.
  • the vertical axis of the spectrogram corresponds to frequency and the horizontal axis corresponds to time.
  • a vertical cross-section of the spectrogram essentially corresponds to one frame and represents the normalized energy spectrum of the frame as a function of frequency.
  • the energy spectrum of a frame is defined as the squared magnitude of the Discrete Fourier Transform of each frame, and always assumes a positive value.
  • the normalized energy spectrum of a frame is obtained by normalizing the energy spectrum of the frame by the total energy in the frame; i.e., the sum of the energy spectrum over all frequencies in the frame.
  • a strong banding structure (i.e., generally horizontal white lines) exists between frame nos. 50 and 350 .
  • the rest is basically noise.
  • the bands are harmonics (multiples) of the pitch frequency and move closer and farther apart as the pitch changes.
  • the dominant band in each frame is not the pitch frequency, but some harmonic of it. Which harmonic is emphasized depends strongly upon the shape of the vocal tract and mouth at the time instant.
  • FIG. 4A There are about ten notes in FIG. 4A with the breakpoints being indicated by the vertical white lines 160 in the image. (Lines 160 are not part of the spectrogram but are merely used to indicate the position of the breakpoints in the image.) Breakpoints between notes can be seen where the dominant band shifts lower because constrictions in the vocal tract reduce the amount of high frequency energy uttered. This is shown more clearly in FIGS. 4B and 4C .
  • FIG. 4B shows the normalized energy spectrum plotted versus frequency for frame no. 150 , which is a near breakpoint.
  • FIG. 4C shows the same kind of plot for frame no. 170 , which is in the middle of a hummed note. A shift in the energy distribution to higher frequencies is clearly evident. Note that these plots are essentially a cross-section through a vertical slice of the spectrogram illustrated in FIG. 4A .
  • the SED indicator 90 represents the shift in energy distribution. This a numerical measure which combines the spectral energies in each frame in such a way that the value of that measure is large if the energy distribution is concentrated in certain frequency bands and small if the energy distribution is concentrated in others. There are a variety of ways to compute the SED indicator.
  • the SED indicator 90 can be computed as the first moment of the energy spectrum in each frame divided by the zero th moment. More particularly, let X(k) be the energy spectrum at frequency bin k; the corresponding spectral energy distribution measure is given by ⁇ k ⁇ kX ⁇ ( k ) ⁇ k ⁇ X ⁇ ( k ) . The summation is carried out over all frequency bins from 0 (DC) up to the frequency bin corresponding to the Nyquist frequency. Frequency bins past the Nyquist frequency contain no additional information due to aliasing. This results in large values if X(k) is concentrated around large frequencies (large k). The graph of FIG.
  • FIG. 5B plots the SED indicator (when computed as just described) for the sample input melody of FIG. 4A , i.e., for all frames.
  • vertical lines 162 indicate the positions of breakpoints.
  • FIG. 5A repeats FIG. 4A to facilitate comparison. Note that the SED indicator drops to minimum values at or near the breakpoints.
  • the breakpoint detection block 100 preferably derives a “minimum measure” at each frame, which is positive if there is a local minimum in the SED indicator “near” the corresponding frame, and zero otherwise.
  • the number of “near” frames is, for example, 15 frames before and 15 frames after the present frame.
  • the amplitude of the minimum measure is larger the “deeper” the local minimum.
  • a linear relationship is preferably employed for determining the amplitude of the minimum measure but other types of relationships can be employed in the alternative such as power, exponential, and logarithmic relationships.
  • FIG. 5C shows an example of the minimum measure for the SED indicator shown in FIG.
  • the breakpoint detection block 100 uses the minimum measure to determine candidate breakpoints by finding the locations of the positive peaks therein.
  • an additional or alternative method for identifying breakpoints is by determining locations of rapid changes in the valid pitch estimate across frames.
  • the confidence measure 110 for each candidate breakpoint is preferably a weighted sum of four numbers.
  • the first number is large if the absolute value of the SED indicator is “small” in the neighborhood of the breakpoint, e.g., less than about 75% of the average value over the input waveform.
  • the second number is large if the minimum measure in the vicinity of the breakpoint is “large”, e.g., larger than about 80% of the maximum value over the input waveform.
  • the third number is large if the rate of change of pitch at the breakpoint is “large”, e.g., more than about 10 semitones per second.
  • the fourth number is large if the average energy in frames around the breakpoint is “small”, e.g. less than 50% of the maximum value in some neighborhood around the candidate breakpoint.
  • each of these numbers is weighted equally, although a variety of weightings may be used in the alternative.
  • a confidence measure 122 is also associated with each note identified between two breakpoints 125 .
  • This confidence measure is designed to indicate the possibility that the identified note does not contain more than one note from the input melody, due to a missed breakpoint in the breakpoint detection block 100 .
  • the note confidence measure 122 is a weighted sum of four numbers. The first number is small if the variation of the SED indicator for frames within the note is “large,” e.g., the difference between the maximum and minimum value is greater than some percentage (e.g. 20%) of the average value. The second number is small if the maximum “minimum measure” taken over all frames in the note is large, e.g. greater than 20% of the maximum value over the input waveform.
  • the third is small if the variation of the identified pitch periods for frames inside the note is large; e.g. the maximum and minimum values vary by more than one semitone.
  • the fourth is small if the variation in the energy level for frames inside the note is “large”; e.g. the difference between the maximum and minimum value is larger than some percentage (e.g. 20%) of the average value.
  • some percentage e.g. 20%
  • the pitch for each note is determined by merging the pitch periods across the frames falling between two breakpoints delineating that note. It is preferred to merge the pitch by finding the median of the pitch estimates between the two breakpoints. The median computation is less sensitive than the average to occasional large errors in the pitch estimates at individual frames. This yields the note pitch 135 .
  • the differential note and timing file 150 is generated by block 140 .
  • the pitch ratio and the beat duration ratio are expressed as the log of the ratio between two consecutive notes which are given as follows:
  • the note-matching engine 16 ( FIG. 1 ) is a score-based engine. It generates a score for each song in the reference database 14 based on the similarity of the input melody input to the songs in the database, taking into account the confidence levels of each identified breakpoint and each extracted note. By using dynamic programming the engine 16 attempts to compensate for errors generated either by the user, who may have vocalized or played the melody with wrong notes or wrong beats, or by the melody-note conversion subsystem 12 , which may miss some notes, over-count notes or measure the note duration incorrectly.
  • the preferred embodiment uses relative beat and note information for the matching process because the user may vocalize or play the melody in any scale, and not necessarily the 12-tone octave scale. Similarly, the user may vocalize or play the melody in any tempo. Therefore, relative pitch and beat data is preferred.
  • the inputs to the matching engine 16 are the differential note and timing file 150 and candidate differential note and timing files from the music database 14 .
  • the note-matching engine 16 employs a dynamic programming approach described below to solve this problem in an optimal manner.
  • the engine 16 sets up a 2-dimensional matrix 180 for each song matching, as exemplified in FIG. 10 .
  • YRF represents the pitch ratio and YRT represents the beat duration ratio of the corresponding entry.
  • Each entry X i is a 4-tuple or vector (XRF i , XRT i , XICON I , XDCON i ) where XRFi, XRT i , XICON i and XDCON i represent the pitch ratio, the beat duration ratio, the confidence level of the note breakpoint, and the confidence level of the note preceding the breakpoint, respectively.
  • the cost of a matching between an entry in X and an entry in Y is defined as the weighted sum of the absolute difference between the corresponding RF and RT.
  • the cost reflects the error of matching X i with Y j . If an entry Xi in X is perfectly matched with an entry Y j in Y, the cost of match is equal to zero. The objective of the song-matching algorithm is to find the subsequence of Y with the minimum matching cost with X. The score of matching is thus the cost of matching. The lower the score, the better the match. If there is no insertion or deletion error in the input differential note and timing file 150 , then the cost of matching the string (X 1 ,X 2 , . . . ,X n ) with a sub-string (Y j , . . .
  • the j index may range from 1 to m ⁇ n+1 (where m is the total number of notes in Y).
  • the lowest value of min_match_cost( ) (as j ranges from m ⁇ n+1) is selected as the score for the candidate song.
  • Insertion and deletion are not the norm. So, the matching process, although it allows for insertion and deletion, also adds a penalty term when the engine 16 tries to match notes assuming there are k insertions or deletions.
  • the conversion subsystem 12 provides a confidence level for every breakpoint and every note that indicates how likely the breakpoint is a “correct” breakpoint and how likely the note is a “correct” note.
  • a low breakpoint confidence level means that the transition is likely to be a wrong transition and hence may result in an insertion error. So a low breakpoint confidence level also implies the note is likely to be an insertion.
  • a low note confidence level means that the note is likely to be composed from several notes and breakpoints are mistakenly deleted.
  • the breakpoint confidence level reflects note insertion error and the note confidence level reflects note deletion error.
  • the penalty should be lowered. For this reason the engine 16 adjusts the penalty by weighting it with the breakpoint and note confidence levels.
  • a breakpoint that is associated with a low confidence level is more likely to be an insertion and hence incurs a lower penalty during matching for note insertion.
  • a note that is associated with a low confidence level is more likely to be a deletion and hence incurs a lower penalty during matching for note deletion.
  • FIG. 10 illustrates the above cost calculation.
  • This figure shows the first 4 ⁇ 4 matrix for a song in a database being compared against a four-note hummed melody.
  • the note matching engine 16 operates in a reverse direction, i.e., the last note of the hummed melody is considered first against the latest notes of the song. For each matrix point, the engine 16 seeks a preceding note having the lowest cost, which translates into highest similarity in relative pitch, relative beat duration and confidence level.
  • the melody retrieval system 10 can be used in, but is not limited to, the following applications:
  • Password protection Rather than having a text-based password protection mechanism to access user accounts and the like, a query based on humming can be employed in the alternative.
  • One aspect of the invention is concerned with estimating or determining breakpoints based on changes in the spectral energy distribution of the input melody.
  • One implementation of the SED indicator has been described. There are alternative ways of computing the SED indicator which nevertheless yield similar properties to the above-described implementation.
  • One broad class of SED indicators is defined by ⁇ k ⁇ f ⁇ ( k ) ⁇ g ⁇ ( X ⁇ ( k ) ) ⁇ ⁇ g ⁇ ( X ⁇ ( k ) ) where f(k) and g(X(k)) are non-negative and non-decreasing functions of k and X(k), respectively.
  • the SED indicator is defined so that it achieves large values if the energy spectrum is concentrated in high frequencies and small values if the energy spectrum is concentrated at low frequencies.
  • An inverse relationship may be employed.
  • alternative embodiments may choose different frequency ranges, such as achieving large values within a band or bands of frequencies and low values outside that band or bands. This might be done to differentiate other types of breakpoints.
  • the SED indicator need not be computed from the energy spectrum.
  • the SED indicator illustrated in FIG. 5A could be computed by estimating the slope at the origin of the auto-correlation of each frame and normalizing that slope by the value at the origin. This is due to the fact that the auto-correlation and the energy spectrum are Fourier Transform pairs, and thus contain the same information.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Acoustics & Sound (AREA)
  • Business, Economics & Management (AREA)
  • Educational Administration (AREA)
  • Educational Technology (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Auxiliary Devices For Music (AREA)
  • Electrophonic Musical Instruments (AREA)
US09/804,069 2000-03-13 2001-03-12 Melody retrieval system Abandoned US20070163425A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US09/804,069 US20070163425A1 (en) 2000-03-13 2001-03-12 Melody retrieval system
US11/953,215 US7919706B2 (en) 2000-03-13 2007-12-10 Melody retrieval system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US18873000P 2000-03-13 2000-03-13
US09/804,069 US20070163425A1 (en) 2000-03-13 2001-03-12 Melody retrieval system

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US11/953,215 Division US7919706B2 (en) 2000-03-13 2007-12-10 Melody retrieval system

Publications (1)

Publication Number Publication Date
US20070163425A1 true US20070163425A1 (en) 2007-07-19

Family

ID=22694300

Family Applications (2)

Application Number Title Priority Date Filing Date
US09/804,069 Abandoned US20070163425A1 (en) 2000-03-13 2001-03-12 Melody retrieval system
US11/953,215 Expired - Fee Related US7919706B2 (en) 2000-03-13 2007-12-10 Melody retrieval system

Family Applications After (1)

Application Number Title Priority Date Filing Date
US11/953,215 Expired - Fee Related US7919706B2 (en) 2000-03-13 2007-12-10 Melody retrieval system

Country Status (3)

Country Link
US (2) US20070163425A1 (fr)
AU (1) AU2001252900A1 (fr)
WO (1) WO2001069575A1 (fr)

Cited By (38)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050192795A1 (en) * 2004-02-26 2005-09-01 Lam Yin H. Identification of the presence of speech in digital audio data
US20060021494A1 (en) * 2002-10-11 2006-02-02 Teo Kok K Method and apparatus for determing musical notes from sounds
US20060095323A1 (en) * 2004-11-03 2006-05-04 Masahiko Muranami Song identification and purchase methodology
US20060137516A1 (en) * 2004-12-24 2006-06-29 Samsung Electronics Co., Ltd. Sound searcher for finding sound media data of specific pattern type and method for operating the same
US20060175409A1 (en) * 2005-02-07 2006-08-10 Sick Ag Code reader
US20070051230A1 (en) * 2005-09-06 2007-03-08 Takashi Hasegawa Information processing system and information processing method
US20070162497A1 (en) * 2003-12-08 2007-07-12 Koninklijke Philips Electronic, N.V. Searching in a melody database
US20070214941A1 (en) * 2006-03-17 2007-09-20 Microsoft Corporation Musical theme searching
US7378588B1 (en) * 2006-09-12 2008-05-27 Chieh Changfan Melody-based music search
US20080210082A1 (en) * 2005-07-22 2008-09-04 Kabushiki Kaisha Kawai Gakki Seisakusho Automatic music transcription apparatus and program
US20080215319A1 (en) * 2007-03-01 2008-09-04 Microsoft Corporation Query by humming for ringtone search and download
US20080282872A1 (en) * 2007-05-17 2008-11-20 Brian Siu-Fung Ma Multifunctional digital music display device
US20090125301A1 (en) * 2007-11-02 2009-05-14 Melodis Inc. Voicing detection modules in a system for automatic transcription of sung or hummed melodies
US20090171485A1 (en) * 2005-06-07 2009-07-02 Matsushita Electric Industrial Co., Ltd. Segmenting a Humming Signal Into Musical Notes
US20090222263A1 (en) * 2005-06-20 2009-09-03 Ivano Salvatore Collotta Method and Apparatus for Transmitting Speech Data To a Remote Device In a Distributed Speech Recognition System
US20090276419A1 (en) * 2008-05-01 2009-11-05 Chacha Search Inc. Method and system for improvement of request processing
US20100030775A1 (en) * 2005-11-10 2010-02-04 Melodis Corporation System And Method For Storing And Retrieving Non-Text-Based Information
US20100082616A1 (en) * 2008-09-30 2010-04-01 Kddi Corporation Apparatus for music retrieval, method for music retrieval, and program for music retrieval
US20110154977A1 (en) * 2009-12-30 2011-06-30 Motorola, Inc. Method and apparatus for best matching an audible query to a set of audible targets
US20120022859A1 (en) * 2009-04-07 2012-01-26 Wen-Hsin Lin Automatic marking method for karaoke vocal accompaniment
CN102522083A (zh) * 2011-11-29 2012-06-27 北京百纳威尔科技有限公司 一种应用移动终端搜索哼唱歌曲的方法及其移动终端
US20120259637A1 (en) * 2011-04-11 2012-10-11 Samsung Electronics Co., Ltd. Method and apparatus for receiving audio
KR101194540B1 (ko) 2010-10-14 2012-10-24 에스케이커뮤니케이션즈 주식회사 음악의 비-텍스트 정보 추출 및 이를 이용한 음악 검색 방법 및 장치
US20130061166A1 (en) * 2011-09-07 2013-03-07 Nhn Corporation Apparatus and method for providing search service
US20130179439A1 (en) * 2001-05-16 2013-07-11 Pandora Media, Inc. Methods and Systems for Utilizing Contextual Feedback to Generate and Modify Playlists
US20140078331A1 (en) * 2012-09-15 2014-03-20 Soundhound, Inc. Method and system for associating sound data with an image
US20140123836A1 (en) * 2012-11-02 2014-05-08 Yakov Vorobyev Musical composition processing system for processing musical composition for energy level and related methods
US9640159B1 (en) 2016-08-25 2017-05-02 Gopro, Inc. Systems and methods for audio based synchronization using sound harmonics
US9653095B1 (en) * 2016-08-30 2017-05-16 Gopro, Inc. Systems and methods for determining a repeatogram in a music composition using audio features
US9697849B1 (en) 2016-07-25 2017-07-04 Gopro, Inc. Systems and methods for audio based synchronization using energy vectors
US9756281B2 (en) 2016-02-05 2017-09-05 Gopro, Inc. Apparatus and method for audio based video synchronization
CN107564540A (zh) * 2017-07-05 2018-01-09 珠海市维想科技有限公司 一种多功能钢琴识别教学辅助系统
WO2018018283A1 (fr) * 2016-07-24 2018-02-01 张鹏华 Procédé de comptage pour conditions d'utilisation d'une technique de reconnaissance d'informations de chanson, et système de reconnaissance
US9916822B1 (en) 2016-10-07 2018-03-13 Gopro, Inc. Systems and methods for audio remixing using repeated segments
US20180276297A1 (en) * 2016-03-24 2018-09-27 Tencent Technology (Shenzhen) Company Limited Audio identification method and apparatus, and computer storage medium
US20200357369A1 (en) * 2018-01-09 2020-11-12 Guangzhou Baiguoyuan Information Technology Co., Ltd. Music classification method and beat point detection method, storage device and computer device
CN113497970A (zh) * 2020-03-19 2021-10-12 字节跳动有限公司 视频处理方法、装置、电子设备及存储介质
CN120388550A (zh) * 2025-06-30 2025-07-29 潍坊学院 一种长笛民族性曲库构建方法及系统

Families Citing this family (29)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE10117870B4 (de) * 2001-04-10 2005-06-09 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Verfahren und Vorrichtung zum Überführen eines Musiksignals in eine Noten-basierte Beschreibung und Verfahren und Vorrichtung zum Referenzieren eines Musiksignals in einer Datenbank
DE10144087B4 (de) * 2001-09-08 2008-10-30 Promediascan Ag Verfahren zur Erkennung und Registrierung urheberrechtlich geschützter Musiksequenzen in Rundfunk- und Fernsehprogrammen
US6747201B2 (en) * 2001-09-26 2004-06-08 The Regents Of The University Of Michigan Method and system for extracting melodic patterns in a musical piece and computer-readable storage medium having a program for executing the method
DE50214167D1 (de) * 2002-07-10 2010-03-04 Palm Inc Verfahren zum Auffinden einer Tonfolge
US7386357B2 (en) * 2002-09-30 2008-06-10 Hewlett-Packard Development Company, L.P. System and method for generating an audio thumbnail of an audio track
JP4302967B2 (ja) * 2002-11-18 2009-07-29 パイオニア株式会社 楽曲検索方法、楽曲検索装置及び楽曲検索プログラム
FR2856817A1 (fr) * 2003-06-25 2004-12-31 France Telecom Procede de traitement d'une sequence sonore, telle qu'un morceau musical
WO2005050615A1 (fr) * 2003-11-21 2005-06-02 Agency For Science, Technology And Research Procede et appareil d'appariement et de representation de melodies pour l'extraction de musiques
US7271329B2 (en) * 2004-05-28 2007-09-18 Electronic Learning Products, Inc. Computer-aided learning system employing a pitch tracking line
US7227074B2 (en) * 2004-09-24 2007-06-05 Microsoft Corporation Transport control for initiating play of dynamically rendered audio content
DE102004049478A1 (de) * 2004-10-11 2006-04-20 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Verfahren und Vorrichtung zur Glättung eines Melodieliniensegments
DE102004049477A1 (de) * 2004-10-11 2006-04-20 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Verfahren und Vorrichtung zur harmonischen Aufbereitung einer Melodielinie
DE102004049517B4 (de) * 2004-10-11 2009-07-16 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Extraktion einer einem Audiosignal zu Grunde liegenden Melodie
US20070276668A1 (en) * 2006-05-23 2007-11-29 Creative Technology Ltd Method and apparatus for accessing an audio file from a collection of audio files using tonal matching
US8208643B2 (en) 2007-06-29 2012-06-26 Tong Zhang Generating music thumbnails and identifying related song structure
EP2239727A1 (fr) * 2009-04-08 2010-10-13 Yamaha Corporation Appareil et programme de performance musicale
US8554348B2 (en) * 2009-07-20 2013-10-08 Apple Inc. Transient detection using a digital audio workstation
TWI426501B (zh) * 2010-11-29 2014-02-11 Inst Information Industry 旋律辨識方法與其裝置
JP5970934B2 (ja) 2011-04-21 2016-08-17 ヤマハ株式会社 楽音発生パターンを示すクエリーを用いて演奏データの検索を行う装置、方法および記録媒体
JP5982980B2 (ja) 2011-04-21 2016-08-31 ヤマハ株式会社 楽音発生パターンを示すクエリーを用いて演奏データの検索を行う装置、方法および記憶媒体
US8212135B1 (en) * 2011-10-19 2012-07-03 Google Inc. Systems and methods for facilitating higher confidence matching by a computer-based melody matching system
GB201202515D0 (en) * 2012-02-14 2012-03-28 Spectral Efficiency Ltd Method for giving feedback on a musical performance
JP2013205830A (ja) * 2012-03-29 2013-10-07 Sony Corp トーン成分検出方法、トーン成分検出装置およびプログラム
JP5799977B2 (ja) 2012-07-18 2015-10-28 ヤマハ株式会社 音符列解析装置
AR092642A1 (es) * 2012-09-24 2015-04-29 Hitlab Inc Metodo y sistema para evaluar usuarios de karaoke
US10452710B2 (en) 2015-09-30 2019-10-22 Microsoft Technology Licensing, Llc Selecting content items based on received term using topic model
CN109979483B (zh) * 2019-03-29 2020-11-03 广州市百果园信息技术有限公司 音频信号的旋律检测方法、装置以及电子设备
JP2024526125A (ja) * 2021-06-15 2024-07-17 エムアイアイアール・オーディオ・テクノロジーズ・インコーポレイテッド 自律神経系の生理学的反応を誘発するのに適した特性を有する音楽のセグメントを識別するためのシステムおよび方法
CN115527514B (zh) * 2022-09-30 2023-11-21 恩平市奥科电子科技有限公司 音乐大数据检索的专业声乐旋律特征提取方法

Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4038503A (en) * 1975-12-29 1977-07-26 Dialog Systems, Inc. Speech recognition apparatus
US4282405A (en) * 1978-11-24 1981-08-04 Nippon Electric Co., Ltd. Speech analyzer comprising circuits for calculating autocorrelation coefficients forwardly and backwardly
US4633748A (en) * 1983-02-27 1987-01-06 Casio Computer Co., Ltd. Electronic musical instrument
US5038658A (en) * 1988-02-29 1991-08-13 Nec Home Electronics Ltd. Method for automatically transcribing music and apparatus therefore
US5565639A (en) * 1993-06-30 1996-10-15 Daewoo Electronics Co., Ltd. Apparatus for giving marks on user's singing ability in karaoke
US5567162A (en) * 1993-11-09 1996-10-22 Daewoo Electronics Co., Ltd. Karaoke system capable of scoring singing of a singer on accompaniment thereof
US5596160A (en) * 1993-11-05 1997-01-21 Yamaha Corporation Performance-information apparatus for analyzing pitch and key-on timing
US5619004A (en) * 1995-06-07 1997-04-08 Virtual Dsp Corporation Method and device for determining the primary pitch of a music signal
US5652596A (en) * 1995-09-22 1997-07-29 Hughes Electronics Scanned antenna system and method
US5952596A (en) * 1997-09-22 1999-09-14 Yamaha Corporation Method of changing tempo and pitch of audio by digital signal processing
US6121530A (en) * 1998-03-19 2000-09-19 Sonoda; Tomonari World Wide Web-based melody retrieval system with thresholds determined by using distribution of pitch and span of notes
US6243672B1 (en) * 1996-09-27 2001-06-05 Sony Corporation Speech encoding/decoding method and apparatus using a pitch reliability measure

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4038503A (en) * 1975-12-29 1977-07-26 Dialog Systems, Inc. Speech recognition apparatus
US4282405A (en) * 1978-11-24 1981-08-04 Nippon Electric Co., Ltd. Speech analyzer comprising circuits for calculating autocorrelation coefficients forwardly and backwardly
US4633748A (en) * 1983-02-27 1987-01-06 Casio Computer Co., Ltd. Electronic musical instrument
US5038658A (en) * 1988-02-29 1991-08-13 Nec Home Electronics Ltd. Method for automatically transcribing music and apparatus therefore
US5565639A (en) * 1993-06-30 1996-10-15 Daewoo Electronics Co., Ltd. Apparatus for giving marks on user's singing ability in karaoke
US5596160A (en) * 1993-11-05 1997-01-21 Yamaha Corporation Performance-information apparatus for analyzing pitch and key-on timing
US5567162A (en) * 1993-11-09 1996-10-22 Daewoo Electronics Co., Ltd. Karaoke system capable of scoring singing of a singer on accompaniment thereof
US5619004A (en) * 1995-06-07 1997-04-08 Virtual Dsp Corporation Method and device for determining the primary pitch of a music signal
US5652596A (en) * 1995-09-22 1997-07-29 Hughes Electronics Scanned antenna system and method
US6243672B1 (en) * 1996-09-27 2001-06-05 Sony Corporation Speech encoding/decoding method and apparatus using a pitch reliability measure
US5952596A (en) * 1997-09-22 1999-09-14 Yamaha Corporation Method of changing tempo and pitch of audio by digital signal processing
US6121530A (en) * 1998-03-19 2000-09-19 Sonoda; Tomonari World Wide Web-based melody retrieval system with thresholds determined by using distribution of pitch and span of notes

Cited By (66)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130179439A1 (en) * 2001-05-16 2013-07-11 Pandora Media, Inc. Methods and Systems for Utilizing Contextual Feedback to Generate and Modify Playlists
US20060021494A1 (en) * 2002-10-11 2006-02-02 Teo Kok K Method and apparatus for determing musical notes from sounds
US7619155B2 (en) * 2002-10-11 2009-11-17 Panasonic Corporation Method and apparatus for determining musical notes from sounds
US20070162497A1 (en) * 2003-12-08 2007-07-12 Koninklijke Philips Electronic, N.V. Searching in a melody database
US8036884B2 (en) * 2004-02-26 2011-10-11 Sony Deutschland Gmbh Identification of the presence of speech in digital audio data
US20050192795A1 (en) * 2004-02-26 2005-09-01 Lam Yin H. Identification of the presence of speech in digital audio data
US20060095323A1 (en) * 2004-11-03 2006-05-04 Masahiko Muranami Song identification and purchase methodology
US20060137516A1 (en) * 2004-12-24 2006-06-29 Samsung Electronics Co., Ltd. Sound searcher for finding sound media data of specific pattern type and method for operating the same
US20060175409A1 (en) * 2005-02-07 2006-08-10 Sick Ag Code reader
US8193436B2 (en) * 2005-06-07 2012-06-05 Matsushita Electric Industrial Co., Ltd. Segmenting a humming signal into musical notes
US20090171485A1 (en) * 2005-06-07 2009-07-02 Matsushita Electric Industrial Co., Ltd. Segmenting a Humming Signal Into Musical Notes
US8494849B2 (en) * 2005-06-20 2013-07-23 Telecom Italia S.P.A. Method and apparatus for transmitting speech data to a remote device in a distributed speech recognition system
US20090222263A1 (en) * 2005-06-20 2009-09-03 Ivano Salvatore Collotta Method and Apparatus for Transmitting Speech Data To a Remote Device In a Distributed Speech Recognition System
US20080210082A1 (en) * 2005-07-22 2008-09-04 Kabushiki Kaisha Kawai Gakki Seisakusho Automatic music transcription apparatus and program
US7507899B2 (en) * 2005-07-22 2009-03-24 Kabushiki Kaisha Kawai Gakki Seisakusho Automatic music transcription apparatus and program
US20070051230A1 (en) * 2005-09-06 2007-03-08 Takashi Hasegawa Information processing system and information processing method
US8041734B2 (en) * 2005-11-10 2011-10-18 Soundhound, Inc. System and method for storing and retrieving non-text-based information
US20120036156A1 (en) * 2005-11-10 2012-02-09 Soundhound, Inc. System and method for storing and retrieving non-text-based information
US20100030775A1 (en) * 2005-11-10 2010-02-04 Melodis Corporation System And Method For Storing And Retrieving Non-Text-Based Information
US9053183B2 (en) * 2005-11-10 2015-06-09 Soundhound, Inc. System and method for storing and retrieving non-text-based information
US7518052B2 (en) * 2006-03-17 2009-04-14 Microsoft Corporation Musical theme searching
US20070214941A1 (en) * 2006-03-17 2007-09-20 Microsoft Corporation Musical theme searching
US7378588B1 (en) * 2006-09-12 2008-05-27 Chieh Changfan Melody-based music search
US20080126304A1 (en) * 2006-09-12 2008-05-29 Chieh Changfan Melody-based music search
US8116746B2 (en) * 2007-03-01 2012-02-14 Microsoft Corporation Technologies for finding ringtones that match a user's hummed rendition
US20080215319A1 (en) * 2007-03-01 2008-09-04 Microsoft Corporation Query by humming for ringtone search and download
US9396257B2 (en) 2007-03-01 2016-07-19 Microsoft Technology Licensing, Llc Query by humming for ringtone search and download
US9794423B2 (en) 2007-03-01 2017-10-17 Microsoft Technology Licensing, Llc Query by humming for ringtone search and download
US7674970B2 (en) * 2007-05-17 2010-03-09 Brian Siu-Fung Ma Multifunctional digital music display device
US20080282872A1 (en) * 2007-05-17 2008-11-20 Brian Siu-Fung Ma Multifunctional digital music display device
US8468014B2 (en) * 2007-11-02 2013-06-18 Soundhound, Inc. Voicing detection modules in a system for automatic transcription of sung or hummed melodies
US20090125301A1 (en) * 2007-11-02 2009-05-14 Melodis Inc. Voicing detection modules in a system for automatic transcription of sung or hummed melodies
US8719256B2 (en) * 2008-05-01 2014-05-06 Chacha Search, Inc Method and system for improvement of request processing
US20090276419A1 (en) * 2008-05-01 2009-11-05 Chacha Search Inc. Method and system for improvement of request processing
US8224846B2 (en) * 2008-09-30 2012-07-17 Kddi Corporation Apparatus for music retrieval, method for music retrieval, and program for music retrieval
US20100082616A1 (en) * 2008-09-30 2010-04-01 Kddi Corporation Apparatus for music retrieval, method for music retrieval, and program for music retrieval
US20120022859A1 (en) * 2009-04-07 2012-01-26 Wen-Hsin Lin Automatic marking method for karaoke vocal accompaniment
US8626497B2 (en) * 2009-04-07 2014-01-07 Wen-Hsin Lin Automatic marking method for karaoke vocal accompaniment
US8049093B2 (en) * 2009-12-30 2011-11-01 Motorola Solutions, Inc. Method and apparatus for best matching an audible query to a set of audible targets
US20110154977A1 (en) * 2009-12-30 2011-06-30 Motorola, Inc. Method and apparatus for best matching an audible query to a set of audible targets
KR101194540B1 (ko) 2010-10-14 2012-10-24 에스케이커뮤니케이션즈 주식회사 음악의 비-텍스트 정보 추출 및 이를 이용한 음악 검색 방법 및 장치
US20120259637A1 (en) * 2011-04-11 2012-10-11 Samsung Electronics Co., Ltd. Method and apparatus for receiving audio
US9122753B2 (en) * 2011-04-11 2015-09-01 Samsung Electronics Co., Ltd. Method and apparatus for retrieving a song by hummed query
US20130061166A1 (en) * 2011-09-07 2013-03-07 Nhn Corporation Apparatus and method for providing search service
US10437457B2 (en) * 2011-09-07 2019-10-08 Nhn Corporation Apparatus and method for providing search service
US9785337B2 (en) * 2011-09-07 2017-10-10 Nhn Corporation Apparatus and method for providing search service
CN102522083A (zh) * 2011-11-29 2012-06-27 北京百纳威尔科技有限公司 一种应用移动终端搜索哼唱歌曲的方法及其移动终端
US20140078331A1 (en) * 2012-09-15 2014-03-20 Soundhound, Inc. Method and system for associating sound data with an image
US20140123836A1 (en) * 2012-11-02 2014-05-08 Yakov Vorobyev Musical composition processing system for processing musical composition for energy level and related methods
US8865993B2 (en) * 2012-11-02 2014-10-21 Mixed In Key Llc Musical composition processing system for processing musical composition for energy level and related methods
US9756281B2 (en) 2016-02-05 2017-09-05 Gopro, Inc. Apparatus and method for audio based video synchronization
US10949462B2 (en) * 2016-03-24 2021-03-16 Tencent Technology (Shenzhen) Company Limited Audio identification method and apparatus, and computer storage medium
US20180276297A1 (en) * 2016-03-24 2018-09-27 Tencent Technology (Shenzhen) Company Limited Audio identification method and apparatus, and computer storage medium
WO2018018283A1 (fr) * 2016-07-24 2018-02-01 张鹏华 Procédé de comptage pour conditions d'utilisation d'une technique de reconnaissance d'informations de chanson, et système de reconnaissance
US10043536B2 (en) 2016-07-25 2018-08-07 Gopro, Inc. Systems and methods for audio based synchronization using energy vectors
US9697849B1 (en) 2016-07-25 2017-07-04 Gopro, Inc. Systems and methods for audio based synchronization using energy vectors
US9972294B1 (en) 2016-08-25 2018-05-15 Gopro, Inc. Systems and methods for audio based synchronization using sound harmonics
US9640159B1 (en) 2016-08-25 2017-05-02 Gopro, Inc. Systems and methods for audio based synchronization using sound harmonics
US10068011B1 (en) * 2016-08-30 2018-09-04 Gopro, Inc. Systems and methods for determining a repeatogram in a music composition using audio features
US9653095B1 (en) * 2016-08-30 2017-05-16 Gopro, Inc. Systems and methods for determining a repeatogram in a music composition using audio features
US9916822B1 (en) 2016-10-07 2018-03-13 Gopro, Inc. Systems and methods for audio remixing using repeated segments
CN107564540A (zh) * 2017-07-05 2018-01-09 珠海市维想科技有限公司 一种多功能钢琴识别教学辅助系统
US20200357369A1 (en) * 2018-01-09 2020-11-12 Guangzhou Baiguoyuan Information Technology Co., Ltd. Music classification method and beat point detection method, storage device and computer device
US11715446B2 (en) * 2018-01-09 2023-08-01 Bigo Technology Pte, Ltd. Music classification method and beat point detection method, storage device and computer device
CN113497970A (zh) * 2020-03-19 2021-10-12 字节跳动有限公司 视频处理方法、装置、电子设备及存储介质
CN120388550A (zh) * 2025-06-30 2025-07-29 潍坊学院 一种长笛民族性曲库构建方法及系统

Also Published As

Publication number Publication date
US7919706B2 (en) 2011-04-05
WO2001069575A1 (fr) 2001-09-20
AU2001252900A1 (en) 2001-09-24
US20080148924A1 (en) 2008-06-26

Similar Documents

Publication Publication Date Title
US7919706B2 (en) Melody retrieval system
JP4392898B2 (ja) 音楽情報処理方法
US10043500B2 (en) Method and apparatus for making music selection based on acoustic features
JP5115966B2 (ja) 楽曲検索システム及び方法並びにそのプログラム
Pauws Musical key extraction from audio.
Burred et al. Hierarchical automatic audio signal classification
Lu et al. Repeating pattern discovery and structure analysis from acoustic music data
JP3433818B2 (ja) 楽曲検索装置
Marolt A mid-level representation for melody-based retrieval in audio collections
US8805657B2 (en) Music searching methods based on human perception
US7031980B2 (en) Music similarity function based on signal analysis
JP4243682B2 (ja) 音楽音響データ中のサビ区間を検出する方法及び装置並びに該方法を実行するためのプログラム
Yang Music database retrieval based on spectral similarity
US20030205124A1 (en) Method and system for retrieving and sequencing music by rhythmic similarity
US20030182118A1 (en) System and method for indexing videos based on speaker distinction
Lu Indexing and retrieval of audio: A survey
Haus et al. An audio front end for query-by-humming systems
Zhu et al. Precise pitch profile feature extraction from musical audio for key detection
KR20060132607A (ko) 멜로디 데이터베이스에서 검색하는 방법
Zhu et al. Music key detection for musical audio
Pollastri A pitch tracking system dedicated to process singing voice for music retrieval
Gao et al. Vocal melody extraction via DNN-based pitch estimation and salience-based pitch refinement
Zhang Semi-automatic approach for music classification
JP3934556B2 (ja) 信号識別子の抽出方法及びその装置、信号識別子からデータベースを作成する方法及びその装置、及び、検索時間領域信号を参照する方法及びその装置
Pollastri Melody-retrieval based on pitch-tracking and string-matching methods

Legal Events

Date Code Title Description
AS Assignment

Owner name: PERCEPTION DIGITAL TECHNOLOGY (BVI) LIMITED, VIRGI

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:TSUI, CHI-YING;SHI, BERTRAM;YUNG, CHI WAI;REEL/FRAME:011847/0450;SIGNING DATES FROM 20010514 TO 20010517

STCB Information on status: application discontinuation

Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION