JP2006126841A - Periodic signal enhancement system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a signal enhancement system which enhances signal components and improves an SNR. <P>SOLUTION: A signal enhancement system (100) improves the understandability of speech or other audio signals. The system reinforces selected parts of the signal, can attenuate selected parts of the signal, and can increase the SNR. The system includes a delay logic (204), an adaptive filter (206), and a signal reinforcement logic (218). The adaptive filter can track and enhance the fundamental frequency and harmonics in the input signal (202). The filter output signals can approximately reproduce the input signal, delayed by an integer multiple of the period of the fundamental frequency of the input signal. The reinforcement logic combines the input signal and the filtered signals to produce an enhanced output signal. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

(Claiming priority)
This application is a continuation-in-part of US patent application Ser. No. 10 / 973,575, filed Oct. 26, 2004, titled Periodic Signal Enhancement System. This application is related to US Patent Application No. (), filed () entitled Periodic Signal Enhancement System.

  The present invention relates to a signal processing system, and more particularly to a system capable of enhancing a periodic signal component.

  Signal processing systems play many roles. The audio signal processing system captures sound very clearly, plays the sound, and transmits the sound to other devices. However, audio systems are susceptible to noise sources that can corrupt, mask, or otherwise adversely affect signal components.

  There are many noise sources. Wind, rain, background noise such as engine noise and electromagnetic interference, and other noise sources can contribute to the noise of signals captured, replayed or transmitted to other systems. When the noise level of speech increases, the clarity decreases.

  Some conventional systems attempt to minimize noisy signals with multiple microphones. The signals from each microphone are smartly combined to limit noise. However, in some applications, multiple microphones cannot be used.

  Other systems use noise filters to selectively attenuate the audio signal. The filter may also indiscriminately remove or minimize the desired signal component.

  There is a need for a system that improves the signal.

(wrap up)
The present invention provides a signal enhancement system that can reinforce signal components and improve the SNR of the signal. This system detects, tracks, and enhances non-stationary periodic signal components of the signal. The periodic signal component can represent vowel sounds or other generated sounds. The system can also detect, track, and attenuate quasi-stationary signal components of the signal.

  The enhancement system includes signal input, delay logic, a partitioned adaptive filter, and signal enhancement logic. The piecewise adaptive filter can track the unsteady fundamental frequency component of the input signal based on the delayed unsteady fundamental frequency component of the input signal. The piecewise adaptive filter outputs a plurality of filtered signals. The filtered signal substantially tracks and increases the frequency content of the signal. The enhancement logic combines the input signal and the filtered signal to produce an enhanced signal. A second adaptive filter can be used to track and suppress the quasi-stationary signal component of the input signal.

The present invention further provides the following means.
(Item 1)
Signal input,
A partitioned delay logic coupled to the signal input;
A piecewise adaptive filter coupled to the piecewise delay logic and having a plurality of adaptive filter outputs;
Filter reinforcement logic coupled to the adaptive filter output;
Gain logic coupled to the filter enhancement logic;
Signal enhancement logic coupled to the signal input and gain logic and having an enhanced signal output;
A signal enhancement system comprising:
(Item 2)
The plurality of filter outputs have a first filter output and a second filter output, and the piecewise adaptive filter comprises:
A first adaptive filter comprising:
A first filter coefficient;
The first filter output;
A first adaptive filter with a first error output;
A second adaptive filter comprising:
A second filter coefficient;
The second filter output;
A second adaptive filter with a second error output;
The first filter coefficient is adjusted based on the first error output, and the second filter coefficient is adjusted based on the second error output;
Item 1. The signal enhancement system according to Item 1.
(Item 3)
The first error output has a first difference between the signal input and the first filter output, and the second error output is between the signal input and the second filter output. Item 3. The signal enhancement system of item 2, having a second difference.
(Item 4)
Delay logic has an M1 sample delay coupled to the first adaptive filter and an M2 sample delay coupled to the second adaptive filter;
Item 3. The signal enhancement system according to Item 2.
(Item 5)
Item 5. The signal enhancement system of item 4, wherein the M2 sample delay is in series with the M1 sample delay.
(Item 6)
Item 5. The signal enhancement system of item 4, wherein the first adaptive filter is a length M1 adaptive filter and the second adaptive filter is a length M2 adaptive filter.
(Item 7)
Item 7. The signal enhancement system of item 6, wherein M1 = M2 or M1 = M2 = 1.
(Item 8)
Item 2. The signal enhancement system of item 1, wherein the delay logic comprises a D sample delay selected to set a maximum adaptive pitch.
(Item 9)
Item 2. The signal enhancement system of item 1, wherein the delay logic comprises an L sample delay selected to set an adaptive pitch range.
(Item 10)
Item 2. The signal enhancement system according to Item 1, wherein the delay logic realizes an adaptive pitch range including a human voice pitch.
(Item 11)
Item 2. The signal enhancement system of item 1, wherein the delay logic implements an adaptive pitch range between about 70 Hz and about 400 Hz.
(Item 12)
A method for enhancing a signal, comprising:
Receiving an input signal including a fundamental frequency;
Delaying the input signal by a plurality of different sample delays to obtain a plurality of differently delayed input signals;
Applying a piecewise adaptive filter comprising a plurality of individual adaptive filters to the plurality of differently delayed input signals;
Generating a filtered output by the piecewise adaptive filter, wherein the filtered output is substantially delayed by an integer multiple of the fundamental frequency;
Generating an error signal for each of the plurality of individual adaptive filters;
Adapting each of the individual adaptive filters based on the error signal of the individual adaptive filter;
Enhancing the input signal with the filtered output;
Including the method.
(Item 13)
Generating an output sum of the plurality of adaptive filters;
Biasing the sum by a gain parameter;
The method according to item 12, further comprising:
(Item 14)
Further including determining a maximum pitch to track;
13. A method according to item 12, wherein delaying the input signal comprises delaying the input signal by D samples, where D is selected according to the maximum pitch.
(Item 15)
Further comprising selecting a pitch tracking range;
15. A method according to item 14, wherein delaying the input signal comprises delaying the input signal by D + L samples where L is selected to set the pitch tracking range.
(Item 16)
16. A method according to item 15, wherein the pitch range includes a human voice pitch.
(Item 17)
16. The method of item 15, wherein the pitch range extends between about 70 Hz and about 400 Hz.
(Item 18)
A machine-readable recording medium;
Computer readable instructions embodied in the machine readable recording medium for performing the method of any of items 12-17;
With a product.

(Summary)
A signal enhancement system improves comprehension of conversations and other audio signals. The system can enhance the selected portion of the signal and attenuate the selected portion of the signal to increase the SNR. The system includes delay logic, a piecewise adaptive filter, and signal enhancement logic. The piecewise adaptive filter tracks and enhances the fundamental frequency and harmonic components of the input signal. The output signal of the piecewise adaptive filter can almost reproduce the input signal delayed by an integral multiple of the fundamental frequency period of the input signal. The enhancement logic combines the input signal and the filtered signal to produce an enhanced output signal.

  Other systems, methods, features, and advantages of the present invention will become apparent or apparent to those of ordinary skill in the art upon reading the following drawings and detailed description. What are all the additional systems, methods, and features and benefits described above? This description is within the scope of the invention and should be protected by the following claims.

  The invention can be better understood with reference to the following drawings and detailed description. The components in the drawings do not necessarily have to be scaled or emphasized instead of illustrating the principles of the invention. Moreover, in the drawings, the same reference numerals denote the same corresponding structural elements throughout the different drawings.

  The enhancement system detects and tracks one or more fundamental frequency components in the signal. The signal enhancement system enhances the tracked frequency component. An augmentation system may improve the clarity of information in a speech signal or other audio signal. The enhanced signal may have an improved signal to noise ratio (SNR).

  In FIG. 1, signal enhancement system 100 may operate with pre-processing logic 102 and post-processing logic 104. The augmentation system 100 can be implemented in hardware and / or software. The enhancement system 100 may include a digital signal processor (DSP). The DSP can execute the instructions to delay the input signal, track the frequency content of the signal, filter the signal, and / or enhance the spectral content in the signal. Alternatively, augmentation system 100 may include separate logic or circuits, separate logic and a mix of processors, or may be distributed across multiple processors or programs.

  Enhancement system 100 may receive input from input source 106. Input source 106 may include a digital signal source or an analog signal source such as microphone 108. Microphone 108 may be coupled to enhancement system 100 via sampling system 110. Sampling system 110 may convert the analog signal sensed by microphone 108 into a digital format at a selected sampling rate.

  The sampling rate can be selected to capture any desired frequency component. For conversation, the sampling rate can be approximately 8 kHz to about 22 kHz. For music, the sampling rate can be approximately 22 kHz to about 44 kHz. Other sampling rates may be used for conversation and / or music.

  Digital signal sources may include communication interface 112, other circuits or logic in the system on which enhancement system 100 is implemented, or other signal sources. When the input source is a digital signal source, the enhancement system 100 can receive the digital signal samples with or without additional preprocessing.

  Signal enhancement system 100 may also be coupled with post-processing logic 104. Post-processing logic 104 may include an audio playback system 114, a digital and / or analog data transmission system 116, or video processing logic 118. Other post-processing logic can also be used.

  Audio playback system 114 may include digital to analog converters, filters, amplifiers, and other circuits or logic. The audio playback system 114 can be a conversation and / or music playback system. Audio playback system 114 may be used in cellular phones, car phones, digital media players / recorders, radios, stereos, portable game devices, or other devices that employ sound playback.

  Video processing system 118 may include circuitry and / or logic that provides visual output. The signal used to prepare the visual output can be enhanced by a process performed by the enhancement system 100. Video processing system 118 may control a television or other entertainment device. Alternatively, video processing system 118 may control a computer monitor or liquid crystal display (LCD).

  Transmission system 116 may comprise a network connection, a digital or analog transmitter, or other transmission circuitry and / or logic. Transmission system 116 may communicate the enhanced signal generated by enhancement system 100 to other devices. For example, in a car phone, the transmission system 116 communicates the augmented signal from the car phone to a reference station or other recipient via a wireless connection, such as a ZigBee, Mobile-Fi, Ultrawideband, Wi-fi, or WiMax network. Can do.

  FIG. 2 shows the augmentation system 100. The enhancement system 100 includes a signal input 202. Signal input 202 carries an input signal that is processed by enhancement system 100. In FIG. 2, the input signal is indicated by the symbol “x”. The input signal can be a time domain sample of the conversation. For ease of explanation, conversation signals will be discussed below. However, the enhancement system 100 may enhance signals having any other range of frequency content in the audible or non-audible range.

  The augmentation system 100 can process quasi-stationary or non-stationary signals. A non-stationary signal can change its frequency and / or amplitude content relatively quickly with time. Speech is an example of a non-stationary signal.

  With few exceptions, even the fundamental frequency content in the talker's voice changes throughout the conversation. The change in fundamental frequency can be approximately 50 percent or more per 100 ms. However, for the human ear, the conversational voice may have a relatively constant pitch.

  A quasi-stationary signal may change its frequency and / or amplitude content less frequently than a non-stationary signal. Non-stationary signals may originate from machine noise, suppressed human speech, or other sound sources. Slowly changing engine noise or AC generator noise are examples of quasi-stationary signals.

  As shown in FIG. 2, the input signal is coupled with delay logic 204. Delay logic 204 delays the input signal. The delay can vary significantly according to the particular implementation of the augmentation system 100. The delay may correspond to the period of the selected maximum pitch. The maximum pitch may be equal to the maximum pitch in the input signal that the enhancement system 100 augments. The maximum pitch can vary greatly according to the shape and nature of the input signal.

  The speech signal may include a fundamental frequency component from approximately 70 Hz to approximately 400 Hz. Male conversations often include a fundamental frequency component between approximately 70 Hz and approximately 200 Hz. Female conversations often include fundamental frequency components between approximately 200 Hz and about 400 Hz. Children's conversations often include a fundamental frequency component between approximately 250 Hz and about 400 Hz.

  The augmentation system 100 can process input signals including both male and female voice conversations separately or simultaneously in duplicate. In these systems, the period of the highest pitch can generally coincide with the period of the fundamental frequency of female voice. The period of the highest pitch can be approximately 1/300 Hz (approximately 3.3 ms), or it can be another pitch period associated with female voice.

  As an alternative, augmentation system 100 may handle male-only conversations. In these implementations, the period of the highest pitch may coincide with the period of the fundamental frequency of male speech. The period of the highest pitch can be approximately 1/150 Hz (approximately 6.6 ms) or it can be another pitch period.

  The delay logic 204 may delay the input signal by the number of signal samples corresponding to the highest pitch period. The number of signal samples is given by

NSS = MPP * _{f s}

The number of "NSS" signal sample here, "MPP" is the maximum pitch period, "f _s" is the sampling rate. Assuming an MPP of about 3.3 ms and a sampling rate of about 8 kHz, NSS = approximately 27 samples. In FIG. 2, NSS corresponds to _ΔF0MAX .

  The delayed input signal is received by filter 206. Filter 206 includes a filter output 208 that conveys the filtered output signal, indicated by the symbol “y” in FIG. Filter 206 may track one or more frequency components in the input signal based on the delayed input signal. Filter 206 can also track the fundamental frequency in the input signal when the pitch changes during a voice conversation.

  Filter 206 may reproduce, copy, approximate, or otherwise include the contents of the tracked frequency in the filtered output signal. Filter 206 may be a finite time impulse response filter (FIR), or other form of digital filter. The coefficients of the filter 206 can be adapted. Filter 206 may be adapted by a normalized least mean square (NLMS) method or other adaptive filter techniques such as recursive least squares (RLS) or proportional LMS. Other tracking logic including other filters may also be used.

Filter 206 may converge to the fundamental frequency in the input signal. The range of the fundamental frequency f ₀ where the filter 206 converges is given as follows.

ここにΔ_{Ｆ０ＭＡＸ}は最高ピッチの周期（サンプル数で表す）、ｆ_ｓはサンプリング周波数、（Ｈｚの単位で表す）、およびＬはフィルタ２０６の長さ（サンプル数で表す）を示す。フィルタ２０６が周波数成分を追跡する周波数範囲を増加または減少させるために、フィルタの長さＬは増加または減少され得る。

Here, Δ _F0MAX is the period of the highest pitch (expressed in the number of samples), f _s is the sampling frequency, (expressed in units of Hz), and L is the length of the filter 206 (expressed in the number of samples). In order to increase or decrease the frequency range over which the filter 206 tracks frequency components, the filter length L can be increased or decreased.

  In the above example, the maximum pitch was approximately 300 Hz and the delay logic 204 implemented a delay of 27 samples. When the filter length L is 64 samples, the filter 206 tracks the contents of the fundamental frequency in the frequency range from approximately 88 Hz to approximately 296 Hz.

フィルタ２０６は時間と共に適応し得る。フィルタ２０６は１つのサンプル毎を根拠にエラー信号「ｅ」を評価することによって迅速に適応し得る。代替案としては、フィルタ２０６はサンプルの塊り（ｂｌｏｃｋ）またはその他の根拠に基づいて適応し得る。

Filter 206 may adapt over time. Filter 206 can quickly adapt by evaluating error signal “e” on a sample-by-sample basis. Alternatively, the filter 206 may be adapted based on a sample block or other basis.

  In adapting, the filter 206 may change its one or more filter coefficients. The filter coefficients can change the response of the filter 206. The filter coefficients adapt the filter 206 so that the filter 206 attempts to minimize the error signal “e”.

  Error estimator 210 may generate an error signal “e”. The error estimator 210 may be an adder, a comparator, or other circuit or logic. The error estimator 210 may compare the input signal “x” with the filtered output signal “y”.

  As filter 206 converges to the fundamental frequency in the input signal, the error signal decreases. As the error signal decreases, the filtered output signal “y” becomes more similar to the input signal “x” delayed by an integer multiple of the fundamental frequency of the signal. The gain control logic 212 is responsive to the error signal.

  Optional gain control logic 212 may include a multiplier 214 and a gain parameter 216. Gain control logic 212 may attenuate, amplify, or perform other conversions on the filtered output signal. FIG. 2 illustrates that gain control logic 212 applies gain “A” to the filtered output signal to generate gain-controlled signal “Ay”.

  The enhancement logic 218 may enhance the content of the frequency in the input signal “x” with the gain control signal “Ay”. Enhancement logic 218 may be an adder or other circuit and / or logic. Enhancement logic 218 may generate an enhanced output signal.

s = x + Ay

When the error signal increases, the gain control logic 212 may decrease the gain “A”. As the gain decreases, the contribution of the filtered output signal to the enhanced output signal may decrease. The relationship between error signal and gain can be continuous, stepped, linear, or non-linear.

  In one implementation, augmentation system 100 sets one or more error thresholds. If the error signal exceeds the upper threshold, the gain control logic 212 may decrease the gain “A” to 0 (zero). If an upper threshold can be set for the input signal, the gain “A” can be set to zero if e> x. If the error signal is less than the lower threshold, gain control logic 212 may increase gain “A” to 1 (one).

  If the error signal exceeds the upper threshold, the filter control logic 220 may reset the filter 206. When the filter 206 is reset, the control logic 220 may zero-out the filter coefficients, reinitialize the filter coefficients, or take other actions. The control logic 220 may also dynamically change the length of the filter, change the delay performed by the delay logic 204, or change other characteristics of the enhancement system 100. The control logic 220 may also modify the augmentation system to adapt to changes in the environment in which the augmentation system 100 is used, and may adapt the augmentation system 100 for new talkers or other purposes.

  The filter control logic 220 may also control how quickly the filter 206 is adapted, the filter is adapted, or may monitor or control other filter characteristics. In the context of a system that augments non-stationary signals, the control logic 220 can expect to quickly change the frequency and amplitude components in the input signal. The control logic 220 may also predict or determine over time that a particular frequency component in the input signal is dominant.

  The control logic 220 may also determine that the input signal has its frequency content, amplitude, or other characteristic changed from what was predicted or determined. In response, control logic 220 may stop filter 206 from trying to adapt to the new signal content, slow down the adaptation, or take other action. The control logic 220 may filter the filter 206 until the characteristics of the input signal return to what was expected, until a predetermined time has elapsed, until a command to release control is given, or until another time or condition is met. Control can be performed.

  The delay logic 204 prevents the filtered output signal from being an exact duplicate of the current input signal “x”. Thus, the filtered output signal can closely track the selected periodicity in the input signal “x”. When the current input signal “x” is augmented with the filtered output signal “y” and the output signal “s” is generated, it is obtained that the periodic signal components are combined in an increasing direction, which is irregular Noise components can be combined in a decreasing direction. Thus, periodic signal components can be enhanced over noise.

  The delay introduced by delay logic 204 and filter 206 can be approximately one cycle of the fundamental frequency component tracked by filter 206. The delay may correspond to a glottal pulse delay with respect to speech such as vowels. When the filtered output signal is added to the input signal, the delay may add the fundamental frequency component in phase or nearly in phase.

  When applied in phase, the gain obtained for the fundamental frequency content in the enhanced output signal can be approximately 6 dB or more. Noise in the input signal and the filtered output signal causes a phase mismatch. When the input signal and the filtered output signal are added, the increase in noise can be less than the increase in enhanced frequency content, for example 3 dB or less. The enhanced output signal may have an increased SNR.

  The input signal that the enhancement system 100 processes may include multiple fundamental frequencies. For example, when two talkers are speaking at the same time, the input signal may include two non-stationary fundamental frequencies. When there are multiple fundamental frequencies, filter 206 adapts and continues to converge to provide a filtered signal “y” that is a delayed version of the input signal. The enhancement logic 218 may enhance one or more fundamental frequencies present in the input signal.

  In FIG. 3, the graph shows the coefficient 300 for the filter 206. The coefficients are plotted with the coefficient number on the horizontal axis and the magnitude on the vertical axis. The coefficient 300 shows the filter 206 adapted for female conversation.

At any given time, the coefficient 300 can be analyzed to determine a quick estimate of the fundamental frequency in the input signal at good temporal resolution. The coefficient 300 starts to reach a peak around the coefficient 304 (fifth filter coefficient), the coefficient 306 (sixth filter coefficient), and the coefficient 308 (seventh filter coefficient). By searching for a coefficient peak or approximate coefficient peak and determining the corresponding coefficient index “c”, _a quick approximation of the fundamental frequency fa can be as follows:

図３において、係数のピークは、第６フィルタ係数３０６である。８ｋＨｚのサンプリングレートおよび２７サンプル遅延を以下のように想定する。

In FIG. 3, the coefficient peak is the sixth filter coefficient 306. Assume a sampling rate of 8 kHz and a 27 sample delay as follows.

図４において、プロットは、男性の声に適応するように、フィルタ２０６のために係数４００を示す。係数のピークは、係数４０２（第３４フィルタ係数）、係数４０４（第３５フィルタ係数）、および係数４０６（第３６フィルタ係数）の近くに現れる。基本周波数の概算は、以下のようである。

In FIG. 4, the plot shows the coefficient 400 for the filter 206 to accommodate the male voice. The peak of the coefficient appears near the coefficient 402 (the 34th filter coefficient), the coefficient 404 (the 35th filter coefficient), and the coefficient 406 (the 36th filter coefficient). The rough estimate of the fundamental frequency is as follows.

制御ロジック２２０は、時の経過によって変化するような、入力信号の基本周波数を含みつつ、入力信号の多くの特徴における履歴データを記憶し得る。制御ロジック２２０は、入力信号の特徴が予期せずに変化するかどうかを決定する目的として、履歴データを検査し得る。制御ロジック２２０は、フィルタ２０６を制御する適応を行使することによって、または他の行動を取ることによって、応答し得る。

The control logic 220 may store historical data on many features of the input signal, including the fundamental frequency of the input signal, as it changes over time. The control logic 220 may examine the historical data for the purpose of determining whether the characteristics of the input signal change unexpectedly. The control logic 220 may respond by exercising the adaptation that controls the filter 206 or by taking other actions.

  FIG. 5 shows a flow diagram of an act that can be taken to augment the periodic signal. The maximum pitch is selected for processing by augmentation system 100 (Act 502). The delay logic 204 may be set to perform a maximum pitch (Act 504) period.

  The range of frequencies at which the enhancement system 100 operates may also be selected (Act 506). The length of the filter 205 can be set to accommodate a range of frequencies (Act 508). The length of the filter can be changed dynamically during filter 206 operation.

  The input signal can be delayed and filtered (Act 510). The enhancement system 100 may generate an error signal and may be individually adapted to the filter 206 (Act 512). The enhancement system 100 may control the gain of the filtered output signal (Act 514).

  The enhancement system 100 may add an input signal and a gain control signal (Act 516). An enhanced output signal can result. The enhancement system 100 may also determine an estimate of the fundamental frequency (Act 518). The enhancement system 100 may utilize the frequency estimate to control the filter 206 (Act 520) and exercise the adaptation.

  FIG. 6 illustrates a multi-stage augmentation system 600. The enhancement system 600 includes a first filter stage 602 and a second filter stage 604. Filter stages 602 and 604 may respond or adapt at different rates.

  The first filter stage 602 can adapt slowly and control the quasi-stationary signal component. Quasi-stationary signal components may be present in the input signal due to relatively consistent background noise, such as engine noise or environmental effects, or other reasons.

  The signal input 606 is connected to the first stage 602. Signal input 606 may be connected to delay logic 608. The delay logic may perform a delay corresponding to a period of maximum quasi-stationary frequency that can be controlled by the first stage 602.

  The maximum quasi-stationary frequency may be selected according to known or foreseeable characteristics of the environment in which the enhancement system 600 is utilized. Filter control logic 610 may dynamically modify the delay to adapt first stage 602 to its environment. Filter control logic 610 may also control quasi-stationary filter 612.

  The filter 612 in the first stage may include signal components that track logic such as NLMS adapted FIR filters or RLS adapted FIR filters. The filter 612 in the first stage can adapt slowly, eg, an NLMS step size greater than 0 and generally less than 0.01 with a sampling rate of 8 kHz and a filter length of 64, Quasi-stationary periodic signals can be attenuated while minimizing degradation. The first stage filtered output 614 may provide a filtered output signal that generally reproduces a quasi-stationary signal component in the input signal.

  The adaptation of the suppression logic 616 and the slow filter may pass non-stationary signal components through the first stage 602 to the second stage 604. On the other hand, suppression logic 616 may suppress quasi-stationary signal components in the input signal. Suppression logic 616 may be implemented as numerical logic that subtracts the filtered output signal from the input signal.

  Quasi-stationary signal content duplicated in the filtered output signal is removed from the input signal. The output signal generated by the first stage 602 can be as follows.

x ₂ = e ₁ = xy ₁

If “e ₁ ” is the output signal of the first stage, “x” is the input signal and “y ₁ ” is the filtered output of the first stage.

The first stage output 618 can be connected to the second stage 604. Second stage 604 may process signal “x ₂ ” using adaptive filter 206. The filter 206 can adapt quickly, for example, an NLMS step size of approximately greater than 0.6 and less than 1.0 with a sampling rate of 8 kHz and a length of 64 filters, The fundamental frequency in the signal can be tracked.

  The second stage 604 may enhance the non-stationary signal component in the first stage output signal. Non-stationary signal components may be present in the input signal as a result of speech, music or other signal sources. Second stage 604 may process the first stage output signal as described above.

  The augmentation system 600 utilizes a first suppression stage 602 followed by a second augmentation stage 604. The enhancement system 600 can be utilized to enhance non-stationary signal content, such as audio content. In an environment that leads to slowly changing signal components, the enhancement system 600 may remove or suppress slowly changing signal components. For example, in a car phone, the first stage 602 may remove or suppress engine noise, road noise, or other noise, while the second stage 604 is like a male or female voice component. In addition, the non-stationary signal component is enhanced.

  The signal enhancement system 100 can enhance periodic signal content, increase SNR, and reduce noise in the input signal. When applied to a speech signal, the enhancement system 100 can enhance the fundamental speech frequency and enhance vowels or other sounds. The enhancement system 100 can enhance other signals whether they are audible or inaudible.

  All delays introduced by delay logic 204 or 608 and filter 206 or 612 may also be approximately an integer (one or more) of the tracked pitch period cycles. The delay due to additional cycles can change the input signal to a greater extent than waiting for one cycle. Adding a longer delay filter signal to the current input signal can create a special effect on the output signal, such as reverberation, while enhancing the fundamental frequency component.

  In FIG. 7, the signal enhancement system 700 includes a segmented adaptive filter 702 as well as a segmented delay logic 704. The partitioned adaptive filter 702 includes a number of adaptive filters and is shown in FIG. 7 as adaptive filter 1 via “i”. Adaptive filters 1, 2, 3, and “i” are individually labeled 706, 708, 710, 712. The output of each adaptive filter may be connected to gain logic 744 that includes a multiplier that applies a fixed or variable gain parameter to the filter output. FIG. 7 shows gain parameters 714, 716, 718, and 720 applied individually to the outputs of filters 706-712. Gain and filter control logic 722 may perform control of gain parameters 714-720, and may also perform filter adaptation for individual filters 706-712, respectively.

One or more gains loaded with the filter output may be added together by the enhancement logic 724 to obtain a loaded amount of filter output “y _SUM ”. The enhancement logic 726 adds the weighted amount of filter output “y _SUM ” to the input signal “x” to create the output signal “s”. The enhancement logic can be an adder or other signal summer. Partitioned delay logic 704 includes a number of serially connected delay blocks, five of which are labeled as delay blocks 728, 730, 732, 734 and 736.

Each of the filters 706-712 is delayed by the segmented delay logic 704 and after determining a single error signal “e” for that filter based on “x” and the output signal “y” of the filter, The input signal “x” is received. For example, the error signal “e ₁ ” for the first adaptive filter 702 is “e” = “x” − “y ₁ ”. Each adaptive filter 706-712 is adapted to try to minimize its single error signal “e ₁ ”.

  The partitioned filter 702 splits the entire signal tracking task across multiple adaptive filters 706-712. Each adaptive filter 706-712 may process and adapt a portion of all impulse responses of the partitioned filter 702. As a result, each adaptive filter 706-712 may have a shorter length (eg, with a number smaller than a tap) than the longer adaptive filter shown in FIG.

  Given an impulse response performed with 120 taps and 6 adaptive filters, each adaptive filter may process 20 (or any number) of all impulse responses. In other implementations, the number of adaptive filter sections in filter 702 is equal to the length of all impulse responses, and therefore each adaptive filter has a length of one. The total length of the partitioned filter 702 can be chosen with respect to the range of frequencies that track the partitioned filter 702, as described above.

  Adaptive filters 706-712 may vary in length depending on the expected fundamental frequency in the input signal. To process a portion of the impulse response at or near the expected fundamental frequency, adaptive filters 706-712 can be shorter, more rapid adaptive filters. Far from the expected fundamental frequency, adaptive filters 706-712 can be longer and slower adaptive filters. Thus, the length of the adaptive filters 706-712 can be selected to provide rapid adaptation at or near the frequency of interest in the input signal.

Each adaptive filter 706-712 individually uses almost no filter coefficient updates. Adaptive filters 706-712 may be updated more quickly than filters in implementations that use longer adaptive filters. Faster filter updates result in all enhanced tracking runs, especially at high frequencies. The increase in all tracking runs gives itself to the rapidly changing tracking fundamental frequency, whether those frequencies are speech or artificially created. Least mean square (LMS) algorithms, recursive least squares (RLS) algorithms, LMS, variants of RLS, or other techniques may be utilized to update a single error signal “e _i ”.

  Delay logic 704 delays arrival of input signal “x” to one or more filters 706-712. FIG. 7 shows that each filter 706-712 is associated with a respective delay. Each delay block 728-736 may perform a delay of any number of signal samples.

  One implementation uses an initial delay of D samples in the first delay block 728. Each subsequent delay logic 730-736 has a respective configurable delay and is shown in FIG. 7 as M1, M2, M3, Mi samples of delay. Delay block 730 feeds first adaptive filter 706, delay block 732 feeds second adaptive filter 708, and third delay block 734 feeds third adaptive filter 710, thus i th The i th delay block 736 feeds the i th filter 712.

  Delays D, M1,. . . , Mi can correspond to the length of the adaptive filter that the delay block feeds (eg, the number of taps), or can be different from the length of the adaptive filter that the delay block feeds. For example, the length of adaptive filter 710 may be M3 taps, and delay block 734 feeding adaptive filter 706 may delay signal samples by M3 samples.

  If the length of the adaptive filter “i” is less than its associated delay Mi, the adaptive filter can converge faster initially. If the length of the adaptive filter “i” is greater than its associated delay Mi, upon convergence, the adaptive filter may experience a smaller mean square error. The filter length and / or delay logic 730-736 may be set according to the execution guidelines of the execution in which the system 700 is used.

Delay D may be selected to set the range of fundamental frequencies that system 700 will accommodate. The fundamental frequency f ₀ or pitch range that the filter 700 converges or adapts is given by:

この場合、Ｌは区分された適応フィルタ７０２の全体の長さであって、たとえば、Ｌ＝Ｍ１＋Ｍ２＋．．．＋Ｍｉであり、ｆ_Ｓは、サンプリングレートである。

In this case, L is the total length of the segmented adaptive filter 702, for example, L = M1 + M2 +. . . + Mi and f _S is the sampling rate.

  The gain and filter control logic 722 may control the gains 714-720 and the filter adaptation at each, ie, each of the filters 706-712. The control techniques described above for filter control 220 can also be used in signal enhancement system 700. The gains 714-720 may be proportional to the signal-to-noise ratio of the input signal “x” or may be set based on the signal. As the SNR decreases, one or more gains 714-720 may increase in an attempt to suppress noise. As the SNR increases, the one or more gains 714-720 can be reduced or set to zero.

Gains 714-720 may be determined as the filter coefficient function of its corresponding adaptive filter. One expression for the gains 714-720 optionally includes a normalization constant “k” where A _i = f (h _i ) / k.

The function f (h _i ) is an adaptive filter coefficient function and can be defined in many ways depending on the desired enhancement. An example of f (h _i ) is given below.

一つの実行において、式（５）は、ｍ＝２およびフィルタ長さ１で用いられる。増加する「ｍ」によって、高調波がより大きく増大され得る。ゲイン７１４〜７２０は、フィルタ係数に加えてまたは代替として、他の情報に基づいて選択され得るかまたは決定され得る。正規化定数「ｋ」は、
ｋ＝ｍａｘ_ｉ（ｆ（ｈ_ｉ））
にしたがって設定され得る。

In one implementation, equation (5) is used with m = 2 and filter length 1. With increasing “m”, the harmonics can be greatly increased. The gains 714-720 may be selected or determined based on other information in addition to or as an alternative to the filter coefficients. The normalization constant “k” is
k = max _i (f (h _i ))
Can be set according to

  Gains 714-720 may be selected or modified (eg, increased) to amplify the effect of the adaptive filter with coefficients that increase or enhance the periodic component of the input signal. Gains 714-720 can also be selected or modified (eg, reduced or set to zero) to reduce or weaken the periodic component of the input signal (generally a negative factor) Reduce or eliminate the effect of the adaptive filter comprising However, the gains 714 to 720 can be set to other methods depending on the size of the filter coefficient. Thus, the enhancement system 700 can set the gains 714-720 to each, and only enhancement occurs in the system 700.

The enhancement logic 726 generates an enhanced output signal “s”:
s = x + A ₁ y ₁ + A ₂ y ₂ + A ₃ y ₃ +. . . + A _i y _i
FIG. 8 shows an augmentation system 800 that provides an alternative to augmentation system 700. Enhancement system 800 replaces each controlled gain 714-720 with gain logic 802 (eg, multiplier and gain parameters). Gain logic 802 biases the sum of the adaptive filter outputs by a gain parameter “A” 804. The enhancement logic 806 may provide a sum of each adaptive filter output.

  The signal “s” produced by the enhancement systems 700 and 800 includes an enhanced fundamental frequency and harmonics of the fundamental frequency, resulting in a clearer audio signal. Each adaptive filter 706-712 in the enhancement system can be updated independently by its error signal, and will adapt faster to the filter and the whole. Partitioning into multiple adaptive filters leads to reduced smearing between adjacent harmonics, better preservation of smaller harmonics (eg, harmonics near noise levels), and less distortion of non-periodic components of the input signal . Furthermore, the enhancement system 700 can enhance even harmonics embedded in the noise to a level above the noise, and better preserve small harmonics. When selecting between runs, the enhancement system 800 has the advantage of reduced complexity and reduced computational requirements, while the enhancement system 700 takes advantage of the gain of each adaptive filter 702-708 and its effect on the output signal. Has the advantage of providing flexibility to control independently.

  FIG. 9 is a comparison of the frequency performance of the signal enhancement systems 200 and 800. Plot 902 shows the performance of signal enhancement system 200 including input signal 904 and output signal 906. Plot 908 shows the performance of the signal enhancement system 800 including the same input signal 904 and the enhanced output signal 910. Plot 908 shows an improved overall tracking response including an improved high frequency response of signal enhancement system 800 compared to signal enhancement system 200. Output signal 910 more closely tracks the high frequency content of output signal 904.

  Plots 902 and 908 also show improved separation between harmonics achieved by enhancement system 800. Plot 902 shows the frequency response gap 912 between the input signal 904 and the enhanced signal 906. A performance plot 908 of the enhancement system 800 shows that the gap is much smaller, as indicated by reference number 914. The output signal 910 improves the separation between harmonics, thereby reducing smearing between harmonics in the output signal 910.

  FIG. 10 is a comparison of the frequency performance of the signal enhancement systems 700 and 800. Plot 1002 shows the performance of signal enhancement system 800 including input signal 1004 and output signal 1006 generated by enhancement system 800. Plot 1008 shows the performance of signal enhancement system 700 including the same input signal 1004 and output signal 1006. Plot 1008 shows improved overall tracking of enhancement system 700 (each with controlled gains 714-720) including smaller enhancements at higher frequencies.

  Examples of the enhanced smaller harmonics 1012, 1014, 1016, 1018 are labeled in FIG. The enhanced harmonics 1012 and 1014 are located at about 3000-3200 Hz in the plot 1012, which are enhanced by the enhancement system 800. Enhancement system 700 provides a greater enhancement of smaller harmonics, as shown by enhanced harmonics 1016 and 1018 in plot 1008.

  FIG. 11 shows an act flow chart 1100 that may be used to augment the periodic signal. The maximum pitch that the augmentation systems 700 and 800 track is selected (Act 1102). The pitch may be selected according to the signal types expected to be encountered and their characteristics (such as male, female, or child voice characteristics). The overall delay performed by the delay blocks 728-736 may be set to the maximum pitch period (Act 1104).

  The frequency range over which the enhancement systems 700, 800 operate may also be selected (Act 1106). The overall filter length of adaptive filters 702-708 may be set to accommodate the frequency range (Act 1108). The filter length, frequency range and maximum pitch can be changed dynamically during operation of the enhancement system.

  The enhancement system partitions the overall impulse response across multiple adaptive filters 702-708 (Act 1110). The adaptive filter can be partitioned into smaller blocks where the magnitude of the impulse response of the relevant fundamental frequency is large. Any adaptive filter 706-712 may process one or more points of the impulse response. Each adaptive filter 706-712 may process the same or different number of points in the impulse response.

  Enhancement systems 700 and 800 receive an input signal (Act 1112). Enhancement systems 700 and 800 filter the input signal using the segmented adaptive filter (Act 1114). The gain selected for each is adapted to the filtered output signal of the respective adaptive filter (Act 1116). The controlled output signal gain is then added. Alternatively, the gain can be adapted to the sum of one or more filtered output signals. The enhancement system 700, 800 adds the input signal and the controlled output signal gain (Act 1118). As a result of the enhanced output signal, the fundamental frequency and harmonic content are enhanced.

  Enhancement systems 700 and 800 may incorporate pitch detection logic 738 that includes pitch estimation output “p” 740. The pitch detection logic 738 may determine the fundamental frequency of the estimation of the signal component of the input signal as described above (Act 1120). An estimate can be obtained based on the analysis of the filter coefficients to each adaptive filter 706-712 to quickly estimate the fundamental frequency. Frequency estimation or other information provides principles to enhancement systems 700 and 800, and adaptive control such as increasing and decreasing adaptation rate, changing filter length, adding or removing filters, other adaptations, etc., filter 702. Over 708 and gain (Act 1122).

  Enhancement systems 700 and 800 may also include audio detection logic 742 that includes audio detection output “v” 744. Voice detection logic 742 may place peaks at filter coefficients above a preselected threshold (eg, background noise level). Such a coefficient may indicate the presence of a periodic frequency component in the input signal. Vowel sounds can raise the coefficient peak above the background noise level, which can be a particularly strong peak. If a peak above the threshold is indicated, the voice detection logic 742 can assert the voice detection output “v” to indicate that the input signal includes a voicing component.

  Voice detection logic 742 may determine the detection measurement. The detection measurement provides an indication of whether speech is being shown in the input signal. The detection measurement may be the sum of the magnitudes of the positive filter coefficients. If the sum exceeds the threshold, the voice detection logic may assert the voice detection output “v” 744.

  Each adaptive filter 702-708 generates the error signal (Act 1124). Accordingly, each adaptive filter 702 to 708 adapts based on the error signal (Act 1126). Enhancement systems 700 and 800 may continue to provide an enhanced output signal during the time of the input signal (Act 1128).

  FIG. 12 shows a plurality of stage enhancement systems 1202 and a plurality of stage enhancement systems 1204. System 1202 includes slowly adapting a filter stage (eg, stage 602) that couples to signal enhancement system 700. The input signal “x” 1206 couples to the slowly adapting filter stage 602, and the signal enhancement system 700 generates an enhanced output signal “s” 1208. The multiple stage enhancement system 1204 generates an enhanced output signal “s” 1210 using a slowly adapting filter stage 602 coupled to the signal enhancement system 800.

  Slowly adapting filter stage 602 may suppress quasi-stationary signal components. The quasi-stationary signal component can be shown in the input signal due to background noise with a slowly varying frequency content. Slowly adapting filter stage 602 may suppress engine noise or environmental effects or other noise sources with relatively slowly changing frequency characteristics. Signal enhancement systems 700 and 800 follow to increase the periodic frequency content as it passes through the slowly adapting filter stage 602 present in the audio signal.

  The signal enhancement system 200, 600, 700, 800 may be implemented in hardware or software or a combination of hardware and software. The augmentation system may take the form of instructions stored on a machine readable medium such as a disk, EPROM, flash card, or other memory. The augmentation system 200, 600, 700, 800 may be incorporated into a communication device, sound system, gaming device, signal processing software, or other device and program. The enhancement systems 200, 600, 700, 800 can pre-process the microphone input signal and increase the SNR of the vowel sound for subsequent processing.

  While various embodiments of the invention have been described, it will be apparent to those skilled in the art that more embodiments and implementations are possible within the scope of the invention. Accordingly, the invention should not be limited without considering the appended claims and their equivalents.

FIG. 1 shows a signal enhancement system with pre-processing and post-processing logic. FIG. 2 shows a single stage signal enhancement system. FIG. 3 shows a plot of filter coefficients for a filter adapted to a female voice. FIG. 4 shows a plot of the filter coefficients of a filter adapted for a male voice. FIG. 5 is a flowchart of signal enhancement. FIG. 6 shows a multi-stage signal enhancement system. FIG. 7 shows a signal enhancement system including a segmented adaptive filter. FIG. 8 shows an alternative implementation of a signal enhancement system that includes a segmented adaptive filter. FIG. 9 shows a comparison of the frequency characteristics of the signal enhancement systems shown in FIGS. FIG. 10 shows a comparison of the frequency characteristics of the signal enhancement systems shown in FIGS. FIG. 11 is a flowchart of signal enhancement. FIG. 12 shows a multi-stage signal enhancement system.

Explanation of symbols

100, 600, 700, 800 Signal enhancement system 202 Signal input 204 Delay logic 206 Filter 210 Error estimator 212 Gain control logic 218 Signal enhancement logic 220 Control logic 702 Piecewise adaptive filter 704 Piecewise delay logic 744 Gain logic

Claims (18)

Signal input,
Piecewise delay logic coupled to the signal input;
A piecewise adaptive filter coupled to the piecewise delay logic and having a plurality of adaptive filter outputs;
Filter enhancement logic coupled to the adaptive filter output;
Gain logic coupled to the filter enhancement logic;
Signal enhancement logic coupled to the signal input and gain logic and having an enhanced signal output;
A signal enhancement system comprising: The plurality of filter outputs have a first filter output and a second filter output, and the piecewise adaptive filter comprises:
A first adaptive filter comprising:
A first filter coefficient;
The first filter output;
A first adaptive filter with a first error output;
A second adaptive filter comprising:
A second filter coefficient;
The second filter output;
A second adaptive filter with a second error output;
The first filter coefficient is adjusted based on the first error output, and the second filter coefficient is adjusted based on the second error output;
The signal enhancement system of claim 1.   The first error output has a first difference between the signal input and the first filter output, and the second error output is between the signal input and the second filter output. The signal enhancement system of claim 2, having a second difference. Delay logic has an M1 sample delay coupled to the first adaptive filter and an M2 sample delay coupled to the second adaptive filter;
The signal enhancement system of claim 2.   The signal enhancement system of claim 4, wherein the M2 sample delay is in series with the M1 sample delay.   5. The signal enhancement system of claim 4, wherein the first adaptive filter is a length M1 adaptive filter and the second adaptive filter is a length M2 adaptive filter.   The signal enhancement system of claim 6, wherein M1 = M2 or M1 = M2 = 1.   The signal enhancement system of claim 1, wherein the delay logic comprises a D sample delay selected to set a maximum adaptive pitch.   The signal enhancement system of claim 1, wherein the delay logic comprises an L sample delay selected to set an adaptive pitch range.   The signal enhancement system of claim 1, wherein the delay logic implements an adaptive pitch range that includes a human voice pitch.   The signal enhancement system of claim 1, wherein the delay logic implements an adaptive pitch range between about 70 Hz and about 400 Hz. A method for enhancing a signal, comprising:
Receiving an input signal including a fundamental frequency;
Delaying the input signal by a plurality of different sample delays to obtain a plurality of differently delayed input signals;
Applying a piecewise adaptive filter comprising a plurality of individual adaptive filters to the plurality of differently delayed input signals;
Generating a filtered output by the piecewise adaptive filter, wherein the filtered output is substantially delayed by an integer multiple of the fundamental frequency;
Generating an error signal for each of the plurality of individual adaptive filters;
Adapting each of the individual adaptive filters based on the error signal of the individual adaptive filter;
Enhancing the input signal with the filtered output;
Including the method. Generating an output sum of the plurality of adaptive filters;
Biasing the sum by a gain parameter;
The method of claim 12 further comprising: Further including determining a maximum pitch to track;
The method of claim 12, wherein delaying the input signal comprises delaying the input signal by D samples, where D is selected according to the maximum pitch. Further comprising selecting a pitch tracking range;
The method of claim 14, wherein delaying the input signal comprises delaying the input signal by D + L samples where L is selected to set the pitch tracking range.   The method of claim 15, wherein the pitch range includes a human voice pitch.   The method of claim 15, wherein the pitch range extends between about 70 Hz and about 400 Hz. A machine-readable recording medium;
Computer-readable instructions for performing the method according to any one of claims 12 to 17, embodied in the machine-readable recording medium;
With a product.

Images