JP6657965B2 - Audio signal processing device, audio signal processing method, and audio signal processing program - Google Patents

Description

  The present invention relates to an audio signal processing device, an audio signal processing method, and an audio signal processing program for determining a state of an audio signal.

  2. Description of the Related Art In recent years, it has been restricted that a driver operates a telephone to make a telephone call or the like in a running automobile, and a hands-free communication system using short-range wireless communication or the like has been introduced.

  In a hands-free communication system, a voice generated by a speaker (near end) during a call propagates from a speaker of a communication partner (far end) and is input to a microphone, and is transmitted via a telephone line or a network. The so-called echo sound signal, which is heard by the person himself, disturbs the call.

  In order to suppress such an echo sound signal, various echo cancellation and echo suppressor techniques have been proposed. However, when the audio signal is suppressed in a state in which both the far end and the near end are uttering, that is, in a so-called double talk state, not only the unnecessary far end echo sound signal but also the necessary near end sound signal is suppressed. Will be done. Therefore, it is necessary to determine whether or not double talk has occurred.

  Here, for the far end voice signal, it is sufficient to determine the presence / absence of the voice signal of the other party, so that a known voice determination technique can be used. On the other hand, for the near end audio signal, it is necessary to determine not only the presence / absence of an audio signal but also whether the audio is a speaker's audio signal or an echo audio signal. Therefore, it is difficult to determine whether or not the near end voice signal includes the voice signal of the speaker by the known voice determination technology.

  Patent Literature 1 discloses a technique for determining the state of a near end sound based on a volume ratio between an audio output signal and an audio input signal.

JP 2007-53512 A

  In a running automobile, there are various kinds of noises, so that it may not be possible to accurately determine the state of the near end sound based on the volume ratio between the sound output signal and the sound input signal.

  An object of the present invention is to provide an audio signal processing device, an audio signal processing method, and an audio signal processing program that can determine the state of a near end audio signal without depending on the volume.

The present invention is a frequency domain conversion unit that generates a plurality of frequency domain information by converting a plurality of input audio signals acquired at different positions into frequency domain information, respectively, among the plurality of frequency domain information, A relative value calculating unit that calculates a relative value of a time frequency component of at least one set of frequency domain information, and the relative value is included in a range specified based on a relative value threshold stored in advance in a storage unit. A signal determining unit that determines whether the input audio signal includes an audio signal component emitted from a predetermined position based on whether the input audio signal includes an audio signal component emitted from the predetermined position. A histogram of the generated relative values is generated, and the frequency indicating the frequency of appearance is included in a distribution range of the relative values where the frequency is equal to or higher than a predetermined frequency. The maximum and minimum values of the relative value to provide Luo Dio signal processing device and a relative value threshold calculating unit that calculates, as the relative value threshold that.

The present invention generates a plurality of frequency domain information by converting a plurality of input audio signals obtained at different positions into frequency domain information, respectively, wherein at least one set of frequency Calculating the relative value of the time frequency component of the area information, and determining whether the input audio data is included in the range specified based on the relative value threshold value stored in advance in the storage unit. Determining whether the signal includes an audio signal component emitted from a predetermined position; generating a histogram of the relative values generated from the input audio signal including the audio signal component emitted from the predetermined position; The relative value threshold and the maximum value and the minimum value of the relative value included in the distribution range of the relative value, wherein the frequency indicating the frequency is equal to or greater than a preset frequency. Providing Oh Dio signal processing method of calculating Te.

The present invention provides a computer that converts a plurality of input audio signals obtained at different positions into frequency domain information, thereby generating a plurality of frequency domain information, and the plurality of frequency domain information. A relative value calculation process of calculating a relative value of a time frequency component of at least one set of frequency domain information, and a range specified based on a relative value threshold stored in advance in the storage unit, the relative value And a signal determination process for determining whether the input audio signal includes an audio signal component emitted from a predetermined position based on whether or not the input audio signal includes an audio signal component emitted from the predetermined position. Generate a histogram of the relative value generated from the audio signal, the frequency indicating the appearance frequency is equal to or more than a preset frequency Providing said relative value maximum and Luo Dio signal processing program the minimum value to execute the relative value threshold value calculation processing for calculating, as the relative value threshold included in the distribution range of the relative value.

  According to the audio signal processing device, the audio signal processing method, and the audio signal processing program of the present invention, it is possible to determine the state of the near end audio signal without depending on the volume.

FIG. 2 is a block diagram illustrating an audio signal processing device according to the first embodiment. 5 is a flowchart of an initial setting process in the audio signal processing device according to the first embodiment. FIG. 5 is a diagram illustrating a first example of a graph illustrating a range specified by a relative value threshold according to the first embodiment; FIG. 4 is a diagram illustrating a second example of a graph indicating a range specified by the relative value threshold according to the first embodiment. FIG. 7 is a diagram illustrating a third example of a graph illustrating a range specified by the relative value threshold according to the first embodiment; 5 is a flowchart of a signal determination process in the audio signal processing device according to the first embodiment. FIG. 13 is a diagram illustrating an example of an amplitude ratio histogram when the omnidirectional microphone according to the second embodiment is used. FIG. 13 is a diagram illustrating an example of a phase difference histogram when the omnidirectional microphone according to the second embodiment is used. FIG. 14 is a diagram illustrating an example of an amplitude ratio histogram when the front directional microphone according to the second embodiment is used. FIG. 13 is a diagram illustrating an example of a phase difference histogram when the front directional microphone according to the second embodiment is used. FIG. 14 is a diagram illustrating a relationship between an example of a histogram according to the second embodiment and a range of a relative value threshold. FIG. 9 is a diagram showing an enlarged part of an example of a histogram according to the second embodiment;

<First Embodiment>
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a block diagram of an audio signal processing device 1 according to the first embodiment. The audio signal processing device 1 acquires an input audio signal for each audio acquisition unit from two or more audio acquisition units (for example, sensors such as microphones) installed at different positions.

  As shown in FIG. 1, the audio signal processing device 1 includes a signal input unit 10, a frequency domain conversion unit 11, a relative value calculation unit 12, a relative value threshold calculation unit 13, a storage unit 14, and a signal determination unit 15.

  The signal input unit 10 and the storage unit 14 are configured by hardware. The frequency domain conversion unit 11, the relative value calculation unit 12, the relative value threshold value calculation unit 13, and the signal determination unit 15 are based on an audio processing program executed by a calculation unit such as a CPU (Central Processing Unit) or a DSP (Digital Signal Processor). Is achieved. In this case, the audio signal processing program is stored in various types of non-transitory computer-readable storage media and supplied to the computer. Further, each component realized by the program may be configured by hardware.

  The signal input unit 10 acquires a plurality of input audio signals Ain (1) to Ain (n) from a plurality of audio acquisition units (not shown). Then, the signal input unit 10 converts the input plurality of input audio signals Ain (1) to Ain (n) into digital signals. If the input audio signal is already a digital value, there is no need for a configuration for converting it to a digital signal. Hereinafter, an example will be described in which two input audio signals, that is, a first input audio signal Ain (1) and a second input audio signal Ain (2), acquired using two audio acquisition units, are processed.

  When the audio signal processing device 1 is used in a car, it is desirable that the microphone be installed at a position in front of a speaker (mainly a driver) such as in a steering wheel, a sun visor, a map lamp, or a dashboard. Also, the two microphones are desirably installed at intervals of less than several centimeters in order to accurately calculate information on the amplitude ratio and the phase difference.

  The frequency domain conversion unit 11 converts the two input audio signals Ain (1) and Ain (2) acquired by the signal input unit 10 into frequency domain information for each input audio signal, and generates two frequency domain information. . Here, the frequency domain conversion unit 11 divides the input audio signal by a window function having a predetermined frame length, and performs processing such as FFT on each of the divided frames to generate frequency domain information.

  Here, τ is a frame number, f is a frequency, n is an input audio signal number, and the frequency domain information is described as Xn (f, τ). The frequency domain converter 11 generates frequency domain information X1 (f, τ) corresponding to the first input audio signal Ain (1), and generates frequency domain information X1 (f, τ) corresponding to the second input audio signal Ain (2). The information X2 (f, τ) is generated. Then, the frequency domain conversion unit 11 supplies the generated frequency domain information to the relative value calculation unit 12.

  The relative value calculating unit 12 calculates, for each frequency domain information generated by the frequency domain transforming unit 11, a relative value between a time frequency component included in one frequency domain information and a time frequency component included in another frequency domain information. Then, two relative values corresponding to two pieces of frequency domain information are calculated.

  Specifically, the relative value calculation unit 12 calculates at least one of the amplitude relative value and the phase relative value as a relative value. The amplitude relative value indicates a ratio between the amplitude component of the time frequency component included in one frequency domain information and the amplitude component of the time frequency component included in another frequency domain information. The phase relative value indicates a difference between a phase component of a time frequency component included in one frequency domain information and a phase component of a time frequency component included in another frequency domain information.

  Here, an example of a method of calculating a relative value in the relative value calculation unit 12 will be described. The relative value calculation unit 12 calculates, for example, an amplitude relative value α (f, τ) between the frequency domain information X1 (f, τ) and the frequency domain information X2 (f, τ) based on Expression (1). Further, the relative value calculation unit 12 calculates a phase relative value δ (f, τ) between the frequency domain information X1 (f, τ) and the frequency domain information X2 (f, τ) based on the equation (2). In the equation (2), Im indicates an imaginary part of a complex number, and ω indicates an angular frequency.

  The relative value calculating unit 12 supplies the relative value to the relative value threshold calculating unit 13 in the initial setting processing. Further, the relative value calculation unit 12 supplies the relative value to the signal determination unit 15 in the signal determination processing. The initial setting process and the signal determination process will be described later.

  When there are three or more audio acquisition units, the relative value calculation unit 12 determines the number of combinations of all input audio signals acquired from each audio acquisition unit (the number of combinations for selecting two input audio signals from n input audio signals). ), A relative value may be calculated, or a relative value may be calculated for a set (for example, only one set) of some input audio signals.

  The relative value threshold calculator 13 calculates a relative value threshold based on the relative value calculated by the relative value calculator 12.

  The relative value threshold calculator 13 causes the storage 14 to store the relative value supplied from the relative value calculator 12. Then, the relative value threshold calculator 13 generates a two-dimensional histogram using the two stored relative values.

  The relative threshold calculator 13 calculates a relative threshold using a two-dimensional histogram. The relative value threshold calculator 13 causes the storage 14 to store the relative value threshold. The relative value threshold will be described later.

  The signal determination unit 15 determines whether the input signal includes a predetermined signal based on the relative value supplied from the relative value calculation unit 12 and the relative value threshold stored in the storage unit 14.

  That is, the signal determination unit 15 determines whether the input signal includes an audio signal component (hereinafter, referred to as a predetermined audio signal component) emitted from a predetermined position (for example, a driver in a car). Do.

  When an audio signal is acquired by a plurality of audio acquisition units installed at different positions, a difference in amplitude or phase among a plurality of input audio signals is biased to a predetermined range. This is because constant amplitude attenuation and phase shift occur based on the positional relationship between the signal source that emits the audio signal and the plurality of audio acquisition units.

  Therefore, the signal determination unit 15 determines whether or not the input signal includes a predetermined audio signal component based on a difference in the amplitude or the phase of the two input audio signals.

  Specifically, the relative value threshold calculating unit 13 causes the storage unit 14 to store, as a relative value threshold, a value specifying a range in which the amplitude ratio or the phase difference between the two input audio signals is biased.

  The signal determination unit 15 determines whether the relative value supplied from the relative value calculation unit 12 is within a range specified by the relative value threshold. The signal determination unit 15 determines whether or not the input signal includes a predetermined audio signal component based on the determination result as to whether or not the relative value is within the range.

  In the signal determination process, the signal determination unit 15 determines whether the signal has a predetermined audio signal component using the relative value threshold stored in the storage unit 14. This relative value threshold can be stored in the storage unit 14 before the apparatus is installed (for example, at the time of shipping the apparatus). However, in order to enable the signal determination unit 15 to perform the determination with high accuracy, it is preferable that the relative value threshold calculation unit 13 calculates the relative value threshold in an environment where the audio signal processing device 1 is installed.

  Next, an initialization process of the audio signal processing device 1 will be described with reference to FIG. The processing described below is the same for the procedure executed by the audio signal processing method and the audio signal processing program.

  In this initial setting process, it is preferable to generate the initial setting audio signal in a state where there are few audio signal components other than the predetermined audio signal component (such as a voice signal of another speaker or a noise signal). This is to make the peak of the two-dimensional histogram generated by the relative value threshold calculator 13 clearer.

  Further, it is preferable that the initial setting audio signal is generated from a sound output unit located at a position where a predetermined audio signal component is generated. This is for reproducing an amplitude relative value and a phase relative value of a predetermined audio signal component and generating an accurate relative value threshold.

  As described above, the initial setting audio signal may be a predetermined audio signal generated by an audio output unit (not shown) or a voice of a driver sitting in a driver's seat. Further, the processing based on the flowchart of FIG. 2 may be started in response to the initial setting audio signal being emitted from the audio output unit.

  In step S10, the signal input unit 10 determines whether there is an input audio signal. If there is an input audio signal (step S10: YES), the signal input unit 10 shifts the processing to step S11.

  That is, in the initial setting process, the audio signal processing device 1 repeatedly performs the processes of steps S11 to S14 during a period in which the input audio signal is continuously input.

  Note that the frequency domain conversion unit 11 may determine the presence or absence of an input audio signal using the frequency domain information. In this case, the determination of the presence or absence of the input audio signal is performed after step S12 and before step S13.

  In step S11, the frequency domain transform unit 11 divides the input audio signal into frames. The frequency domain conversion unit 11 converts the input audio signal from time domain information to frequency domain information in step S12.

  In step S13, the relative value calculation unit 12 calculates a relative value using a plurality of input audio signals. Then, the relative value calculation unit 12 causes the storage unit 14 to store the relative value calculated in step S14.

  Returning to step S10, when there is no input audio signal input to the signal input unit 10 (step S10: NO), the signal input unit 10 shifts the processing to step S15.

  In step S15, the relative value threshold calculator 13 generates a two-dimensional histogram from the relative values stored in the storage 14.

  Here, the relationship between the two-dimensional histogram and the relative value threshold will be described in detail. FIGS. 3 to 5 show first to third examples of graphs showing ranges determined by the relative value threshold.

  Each of the graphs shown in FIGS. 3 to 5 is a graph showing a two-dimensional histogram in which the horizontal axis represents the amplitude relative value α, the vertical axis represents the phase relative value δ, and the axis in the z direction represents the frequency indicating the appearance frequency of the relative value. is there. That is, each coordinate of the two-dimensional histogram is represented as (amplitude relative value α, phase relative value δ, frequency z) as a frequency corresponding to a set of relative values of α and δ. A coordinate having a high frequency indicates that the frequency of the input audio signal corresponding to the coordinate is high. The hatched portion of each graph indicates a range in which a relative value equal to or higher than a predetermined frequency exists. The isosceles triangle in each graph schematically represents a part of a relative value equal to or higher than a predetermined frequency. Details of each graph will be described later.

  Note that, instead of the frequency, a sum of the amplitude value or the energy value of the time frequency component of the input audio signal corresponding to the coordinates (hereinafter, referred to as frequency) may be used. This is because a time frequency component having a large amplitude value or energy value has a high probability of being a time frequency component derived from a predetermined audio signal component. Not limited to the sum, a statistical value such as an average value or a median value of amplitude values or energy values may be used.

  Returning to FIG. 2, in step S16, the relative value threshold calculating unit 13 detects a range in which a set of relative values whose frequency is equal to or more than a predetermined value exists. The relative value threshold calculator 13 calculates a relative value threshold that specifies this range as a range in which the predetermined audio signal component exists.

  Here, the process in which the relative value threshold calculator 13 calculates the relative value threshold will be described in detail. In the following description, a peak indicating a frequency equal to or higher than a predetermined value is referred to as a significant peak. As the predetermined value, a statistical value such as an average value or a median of the frequencies of each coordinate may be used. The predetermined value may be a value located between the statistical value and the maximum frequency value, a sum of the statistical value and a value indicating a variation such as a standard deviation, or a value obtained by multiplying the maximum frequency value by a predetermined coefficient. May be used.

  In the first example shown in FIG. 3, the upper limit value of the amplitude relative value in the range where a significant peak exists on the two-dimensional histogram is defined as the amplitude ratio upper limit value maxα, and the lower limit value is specified as the amplitude ratio lower limit value minα. . Further, the range is specified by defining the upper limit of the phase relative value in the range where a significant peak exists on the two-dimensional histogram as the phase difference upper limit value maxδ and the lower limit as the phase difference lower limit value minδ.

In the first example shown in FIG. 3, the relative value threshold calculator 13 calculates a rectangular area defined by four values of an amplitude ratio upper limit value maxα, an amplitude ratio lower limit value minα, a phase difference upper limit value maxδ, and a phase difference lower limit value minδ. It is specified as a range where the predetermined audio signal component exists. In this case, the relative value threshold calculation unit 13 causes the storage unit 14 to store four values of the amplitude ratio upper limit value maxα, the amplitude ratio lower limit value minα, the phase difference upper limit value maxδ, and the phase difference lower limit value minδ as relative value threshold values. .

  In the second example illustrated in FIG. 4, the relative value threshold calculator 13 calculates a region of a predetermined shape (for example, an ellipse or a circle) including a part or all of the relative values at which a significant peak exists on the two-dimensional histogram. Is specified as a range in which the predetermined audio signal component exists. In this case, the relative value threshold calculating unit 13 causes the storage unit 14 to store information (for example, the center and radius of a circle, the center of an ellipse, the minor axis and the major axis, and the like) specifying a predetermined shape as the relative value threshold. .

  In the third example illustrated in FIG. 5, the relative value threshold value calculation unit 13 calculates an arbitrary shape including a part or all of the relative value at which a significant peak exists on the two-dimensional histogram (for example, the coordinates of the significant peak are calculated. An area having a shape connected by a straight line or a curve) is specified as a range in which a predetermined audio signal component exists. In this case, the relative value threshold calculating unit 13 causes the storage unit 14 to store part or all of the coordinates of a significant peak located on the outer periphery of an arbitrary shape as the relative value threshold.

  As described above, the relative value threshold value calculation unit 13 specifies a region by a predetermined shape such as a rectangle or an ellipse or an arbitrary shape so as to include all or some of the relative values at which a significant peak is detected. The relative value threshold calculator 13 may use a shape determined in advance, or may select and use a shape most similar to the distribution of significant peaks.

  Next, a signal determination process of the audio signal processing device 1 according to the first embodiment will be described with reference to FIG.

  In step S20, the frequency domain transforming unit 11 divides the input audio signal into frames. In step S21, the frequency domain conversion unit 11 converts the input audio signal from time domain information to frequency domain information.

  In step S22, the relative value calculation unit 12 calculates a relative value using a plurality of input audio signals.

  In step S23, the signal determination unit 15 determines whether the input audio signal includes a predetermined audio signal component with reference to the relative value threshold stored in the storage unit 14.

  The signal determination unit 15 preferably determines whether or not the input audio signal includes a predetermined audio signal component based on whether or not the frequency or the like in the area of the histogram specified by the relative value threshold exceeds a predetermined value. . Further, the signal determination processing may be performed by comparing frequencies and the like inside and outside the area of the histogram specified by the relative value threshold.

  The audio signal processing apparatus 1 according to the first embodiment includes an input audio signal including a predetermined audio signal component based on a relative value threshold stored in advance and a relative value calculated from a continuously input input audio signal. It is determined whether or not.

  As described above, the audio signal processing device 1 according to the first embodiment determines whether the input audio signal includes a signal derived from the predetermined audio signal component based on the relative value calculated from the plurality of input audio signals. I do. By using the relative value, the audio signal processing device 1 of the first embodiment can make the determination without depending on the volume of the input audio signal. Therefore, the audio signal processing device 1 according to the first embodiment can determine the state of the near end audio signal without depending on the volume of the input audio signal.

  The audio signal processing device 1 according to the first embodiment can determine the presence or absence of the far end audio signal using a known audio determination technology, and can determine whether or not double talk has occurred.

  It should be noted that the present invention is not limited to the above-described embodiment, and can be appropriately changed without departing from the gist. For example, a plurality of positional relationships between the signal source of the predetermined audio signal component and the audio acquisition unit may be determined, and the storage unit 14 may store a plurality of relative value thresholds for each of the plurality of positional relationships. In this case, a relative value threshold to be used can be selected from a plurality of relative value thresholds and used according to the usage status of the audio signal processing device 1.

<Embodiment 2>
In the audio signal processing device according to the first embodiment, it is desirable that a so-called omnidirectional microphone having flat directional characteristics is employed as the signal input unit 10. However, due to the configuration of the device, a microphone having a constant directional characteristic or a low-cost microphone may be necessarily used.

  In the former case, it may not be possible to calculate an accurate relative value between two microphones for a signal component arriving from a certain direction due to its structure. In the latter case, there may be individual differences in hardware performance such as input sensitivity, and an accurate relative value may not be calculated.

  In such a case, the distribution of the two-dimensional histogram of the amplitude ratio and the phase difference calculated from the plurality of microphones is more messy than in the case of the non-directional microphone. As a result, it is difficult to determine an appropriate range of the amplitude ratio and the phase difference and calculate the relative value threshold.

  Therefore, in the second embodiment, instead of a two-dimensional histogram of the phase difference and the amplitude ratio, one-dimensional histograms of the phase difference and the amplitude ratio are separately generated, and the distribution range of the amplitude ratio and the phase difference is obtained.

  The difference between the first embodiment and the second embodiment is only the processing of the relative value threshold value calculation unit 13 in the initial setting processing. Therefore, in the second embodiment, only the details of the processing will be described, and the processing of the other blocks will not be described.

  The relative value threshold calculator 13 generates a histogram for each of the amplitude ratio and the phase difference calculated by the relative value calculator 12. In the histogram, it is desirable that the horizontal axis is an amplitude ratio or a phase difference, and the vertical axis is a frequency or the like. In addition, the histogram may have a large variation in the frequency and the like depending on the setting of the class width for calculating the frequency and the like. It is desirable to generate a histogram with the class width being widened to a certain extent, or to perform a smoothing process after the generation of the histogram.

  Here, the difference between the histograms of the omnidirectional microphone and the forward directional microphone will be described. The front directional microphone is an example of a microphone having directivity. 7 and 8 show an example of a histogram when an omnidirectional microphone is used. In each case, the dotted line is a value obtained by smoothing the value such as the actual frequency by a moving average, and the solid line is a value obtained by smoothing the value such as the frequency. The data used for generating the histogram is data when a speaker is located in front of two microphones.

  FIG. 7 shows an example of a histogram in which the horizontal axis represents the amplitude ratio and the vertical axis represents the sum of the amplitude values. A histogram peak appears near the origin (α = 1). In other words, this indicates that the speakers as sound sources are located at the same distance from the two microphones.

  FIG. 8 shows an example of a histogram in which the horizontal axis represents the phase difference and the vertical axis represents the sum of the amplitude values. The peak of the histogram appears near the origin (assuming δ = 0). In other words, this indicates that the speaker as the sound source is located at the same distance from the two microphones.

  When a speaker is positioned in front of the omnidirectional microphone, both histograms show a steep peak near the origin as shown in FIGS. 7 and 8. In this case, the two-dimensional histogram shows a steep peak having a substantially conical shape. As described above, in the case of the omnidirectional microphone, it is easy to obtain the distribution range.

  9 and 10 show an example of a histogram when a microphone having forward directivity is used. As in FIGS. 7 and 8, the dotted line is a value obtained by smoothing a value such as the actual frequency by a moving average, and the solid line is a value obtained by moving average. The data used for generating the histogram is data when a speaker is positioned diagonally forward of the two microphones. That is, the distances from the two microphones to the speaker as the sound source are different.

  FIG. 9 shows an example of a histogram in which the horizontal axis represents the amplitude ratio and the vertical axis represents the sum of the amplitude values. The peak is located away from the origin (α = 1), and coincides with the positional relationship between the speaker as the sound source and the microphone. Further, as compared with the distribution of the omnidirectional microphone, it can be seen that the distribution range is wide and the amplitude value width is wide. Comparing the two distributions, the distribution of microphones having directivity is more random.

  FIG. 10 shows an example of a histogram in which the horizontal axis represents the phase difference and the vertical axis represents the sum of the amplitude values. The peak is located away from the origin (δ = 0), and matches the positional relationship between the speaker as the sound source and the microphone. Further, when compared with the distribution of the omnidirectional microphone, it can be seen that there are a plurality of small peaks.

  In this case, unlike an omnidirectional microphone, the two-dimensional histogram does not have a conical distribution, but is distributed in a mountain-like manner, or has a distribution in which small peaks are scattered. The distribution is messy. In such a random distribution, conditions for determining whether or not a certain relative value is within the range of the relative value threshold are increased, which leads to an increase in processing time.

  After generating the histogram, the relative value threshold calculator 13 calculates a relative value threshold from the distribution range of the relative value for each of the amplitude ratio and the phase difference. Specifically, the maximum value and the minimum value of the relative values included in the distribution range of the relative values in which the frequency indicating the appearance frequency of the relative value is equal to or more than a predetermined value are set as the relative value threshold. That is, as shown in FIG. 11, the maximum value (upper limit value) and the minimum value (lower limit value) of the range of the double arrow, which is the distribution range of the relative value in which the frequency or the like becomes a predetermined value or more, are set as the relative value threshold.

  In the case of a vehicle interior, when a signal component reflected by a side window or the like is delayed and a peak is generated at a position distant from a peak at which the frequency or the like has a maximum value, and a plurality of ranges in which the frequency or the like becomes a predetermined value or more occur. There is also. Since it is necessary to exclude the range caused by the delayed signal component, it is desirable that the range for calculating the relative value threshold is a range including the position of the peak at which the frequency or the like shows the maximum value.

  Here, a method of calculating a predetermined value such as a frequency will be described. The upper limit and the lower limit of the relative value threshold substantially coincide with a point that sharply increases toward the peak, as indicated by a dotted circle in the enlarged view of FIG. This is because a time-frequency component derived from the predetermined audio signal component appears, and the amplitude value sharply increases.

  However, since the portions asymptotic to the x-axis at both left and right ends of the histogram have large fluctuations as shown in an enlarged view in FIG. 12, the point may not be calculated simply by the increased amount of the histogram. Therefore, as a method of detecting the point, there is a method of performing smoothing differentiation of the histogram to obtain an extreme value of the differential value.

  Further, there is a method of calculating a statistical value such as an average value from a histogram and setting the statistical value to the predetermined value. However, when a statistic such as an average value is simply calculated from the entire section of the histogram, the value of a value larger than a desired predetermined value is large because the difference between the values asymptotic to the x-axis at the left and right ends of the histogram and the portion around the peak is large. Will be calculated. That is, the range of the relative value threshold is calculated to be narrow.

  Therefore, first, a provisional predetermined value (for example, a value obtained by multiplying the maximum value by 0.1) is calculated once based on the value of the frequency of the peak indicating the maximum value, and a section having a value equal to or more than that value is excluded. Next, there is a method of calculating a statistical value such as an average value from the remaining sections to obtain the predetermined value. As described above, by calculating the predetermined value stepwise, the section based on the predetermined value is a section that is always larger than the asymptotic portion of the x-axis in FIG. I can catch it well.

  After calculating the relative value threshold value by the above-described processing, the relative value threshold value calculation unit 13 stores the relative value threshold value in the storage unit 14 and completes the initial setting processing. Subsequent signal determination processing is the same as in the first embodiment.

  By the above processing, even if a microphone other than the omnidirectional microphone is used, the processing time does not become enormous, and the relative value threshold that can maintain a certain determination accuracy can be calculated.

Reference Signs List 1 audio signal processing device 10 signal input unit 11 frequency domain conversion unit 12 relative value calculation unit 13 relative value threshold calculation unit 14 storage unit 15 signal determination unit

Claims (8)

A plurality of input audio signals obtained at different positions, by converting each to frequency domain information, a frequency domain conversion unit that generates a plurality of frequency domain information,
Of the plurality of frequency domain information, a relative value calculation unit that calculates a relative value of a time frequency component of at least one set of frequency domain information,
The input audio signal includes an audio signal component emitted from a predetermined position based on whether the relative value is included in a range specified based on the relative value threshold stored in advance in the storage unit. A signal determination unit for determining whether or not
A histogram of the relative value generated from an input audio signal including an audio signal component emitted from the predetermined position is generated, and a frequency indicating an appearance frequency is included in a distribution range of the relative value in which the frequency is equal to or higher than a predetermined frequency. A relative value threshold calculating unit that calculates the maximum value and the minimum value of the relative values as the relative value threshold,
Luo Dio signal processing apparatus comprising a. The relative value threshold calculating unit, when the distribution range is more produced, billed the maximum and minimum values of the relative values included in the distribution range including the maximum value of the power you and the relative value threshold Item 2. The audio signal processing device according to item 1 . The relative value threshold calculating unit is configured to calculate the amplitude or energy of the input audio signal, an audio signal processing apparatus according to 請 Motomeko 2 that generates a histogram of the relative value by using the amplitude value or energy. The relative value calculating unit, and the amplitude relative value indicating the ratio of the amplitude component in the time-frequency components, 請 Motomeko calculate at least one as the relative value of the phase relative values of differences between the phase components of the time-frequency component The audio signal processing device according to any one of claims 1 to 3 . By converting a plurality of input audio signals obtained at different positions into frequency domain information, to generate a plurality of frequency domain information,
Among the plurality of frequency domain information, a relative value of a time frequency component of at least one set of frequency domain information is calculated,
The input audio signal includes an audio signal component emitted from a predetermined position based on whether the relative value is included in a range specified based on the relative value threshold stored in advance in the storage unit. whether or not it is determined,
A histogram of the relative value generated from an input audio signal including an audio signal component emitted from the predetermined position is generated, and a frequency indicating an appearance frequency is included in a distribution range of the relative value in which the frequency is equal to or higher than a predetermined frequency. Calculating the maximum value and the minimum value of the relative value to be used as the relative value threshold
Oh Dio signal processing method. 6. The audio signal processing method according to claim 5, wherein at least one of an amplitude relative value indicating a ratio of amplitude components of the time frequency component and a phase relative value indicating a difference between phase components of the time frequency component is calculated as the relative value. On the computer,
A plurality of input audio signals obtained at different positions, by converting each into frequency domain information, frequency domain conversion processing to generate a plurality of frequency domain information,
Among the plurality of frequency domain information, a relative value calculation process of calculating a relative value of a time frequency component of at least one set of frequency domain information,
The input audio signal includes an audio signal component emitted from a predetermined position based on whether the relative value is included in a range specified based on the relative value threshold stored in advance in the storage unit. Signal determination processing for determining whether or not
A histogram of the relative value generated from an input audio signal including an audio signal component emitted from the predetermined position is generated, and a frequency indicating an appearance frequency is included in a distribution range of the relative value in which the frequency is equal to or higher than a predetermined frequency. Relative value threshold calculation processing for calculating the maximum value and the minimum value of the relative values as the relative value threshold,
Allowed to run Luo Dio signal processing program. In the relative value calculation process, a process of calculating at least one of an amplitude relative value indicating a ratio of an amplitude component of a time frequency component and a phase relative value indicating a difference between phase components of the time frequency component as the relative value is executed. An audio signal processing program according to claim 7.

Images