JP2010217268A - Low delay signal processor generating signal for both ears enabling perception of direction of sound source - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device capable of determining frequency of a boundary for separating bands and reducing volume of processings for a specific frequency band in acoustic signal processings. <P>SOLUTION: In this low delay signal processor, volume of processings in the band affecting no perception of the direction of sound source in signal analysis processings and signal generation processings is reduced or omitted based on the characteristics of perception of the direction of sound source of the sense of hearing being strong on deterioration of signal with respect to the specific frequency band. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

      The present invention enables generation of a signal capable of direction perception while reducing the processing by limiting the frequency processing band in signal analysis or signal generation processing in a system that presents an acoustic signal to both ears. The present invention relates to a low delay signal processing apparatus.

      The sound signal incident on both ears physically includes various reflected sounds in a sound propagation space such as a room in addition to the direct sound from the sound source. Even in an anechoic space where there is no reflected sound, there are reflected sounds on the listener's own head, auricle, and trunk. Since the frequency characteristics of these direct and reflected sounds depend on the direction of the sound source, it is said that it can be used as a clue for direction perception by hearing the slight difference.

      Furthermore, the signal difference between both ears is also an important clue for direction perception. As a typical clue, an interaural time difference, an interaural phase difference, and an interaural level difference of an incident sound due to a path difference from a sound source to both ears are known.

              In general, when a sound signal is presented to both ears using headphones or the like, a clue for perception of the direction cannot be obtained, and an in-head localization that feels that the sound can be heard from the inside of the head occurs. Therefore, a method of realizing a virtual sound source by simulating a clue by signal processing is known.

          In order to realize a system that suppresses ambient noise and emphasizes sound in a specific direction using a binaural sound signal presentation device, in order to avoid the above-mentioned localization in the head, sound direction perception Therefore, it is necessary to detect the direction of sound arrival in a way that does not impair the sound, and to appropriately generate a signal capable of perceiving the direction.

      Various devices and systems for estimating a sound source in a specific direction using a plurality of microphones and separating or emphasizing the sound source have been disclosed.

              Patent Document 1 discloses an apparatus that emphasizes a target sound component using a weight of a cross-correlation coefficient of sound signals received by a plurality of microphones.

          In Patent Document 2, the sound source direction is estimated by forming a sound collecting beam in a plurality of directions using a plurality of microphones, and then the direction area is divided into small areas to further form a sound collecting beam. A sound collecting device with improved performance is disclosed.

          Patent Document 3 discloses a sound source direction detection device and a sound source direction detection method that apply temporal frequency analysis such as an inner ear model to sound signals received by both pseudo-ears.

      Patent Document 4 discloses an apparatus that divides an input sound signal into frequency bands using a digital filter, estimates and enhances the direction of arrival of the sound signal, and presents the sound signal to both ears.

JP2004-289762 JP2007-13400 JP2007-85734 JP 2004-325284 A

      However, Patent Documents 1 and 2 have one output channel, and a clue for direction perception is lost. Patent Document 3 is an invention relating to sound source direction detection by a machine that does not include a sound signal generation device and that simulates human perception processing.

      The means for dividing each frequency band in Patent Document 4 has a problem that the required filter coefficient length differs depending on the frequency band when fine control of the low frequency band is performed. In particular, in order to realize a filter having desired performance in a low frequency range, a length of several thousand to several tens of thousands is required, and a processing delay of several tens to several hundreds of milliseconds occurs. This delay is sufficiently perceived, and is a problem to be solved in order to realize a system that operates in real time.

      In order to solve the above problems, the present invention provides an index of a processing band that can be reduced in the analysis and generation processing of an input signal based on an experimental discovery that the frequency characteristics of a specific frequency band do not affect the direction perception of the sound source. An object of the present invention is to provide a low-delay signal processing apparatus that generates a sound signal that does not affect direction perception and presents it to both ears.

      The present invention is a device that estimates the direction of a sound source from sound signals received by a plurality of microphones and presents it to both ears after performing signal processing such as enhancement processing, but performs processing so as not to affect cues for direction perception Therefore, it is desirable to perform in a manner corresponding to each ear, and there are two microphones.

    The present invention includes a signal analyzer for direction detection that divides an input sound signal into bands by frequency analysis and identifies the direction of arrival, and a signal generator that performs acoustic signal processing such as speech enhancement. Applying different precision filters provides reduced processing and reduced delay.

      In order to estimate the sound source direction from the input signals of the two channels, an interaural level difference (ILD) and an interaural phase difference (IPD), which are cues for direction perception, are calculated. Since ILD and IPD have frequency dependency, they are divided into frequency bands by a band pass filter. These processes are performed based on the apparatus described in Patent Document 4.

      In the signal analysis process and the signal generation process, the present invention realizes a reduction in group delay and a reduction in processing time caused by the bandpass filter process by limiting the band to be applied. The width of the processing band is determined based on a direction perception experiment.

      According to the configuration of the present invention, it is possible to generate a signal that controls or omits processing of a specific frequency band and does not affect sound source direction perception, and only reduces physical group delay. In addition, it is possible to reduce the overall processing load and reduce power consumption.

It is a figure explaining the Example of this invention. It is a figure explaining a sound source direction estimation part. It is a figure explaining a signal generation processing part. It is a figure explaining the relationship between a processing zone | band and filter coefficient length. It is a figure of the frequency amplitude characteristic shown as an example of a head related transfer function. It is a figure which shows an example of the result of the direction localization experiment implemented in order to determine the boundary of the low region and high region in this invention.

      The best mode for carrying out the present invention will be described below with reference to the drawings based on the embodiments.

      The embodiment of the present invention shows an example in which knowledge obtained experimentally regarding the influence on the perception of the sound source direction in the low frequency range is applied.

    FIG. 1 is a diagram showing a configuration of an embodiment of the present invention. The system is mainly composed of two blocks, which are a signal analysis process and a signal generation process, respectively. In the embodiment, it is assumed that the sound source direction is estimated and signal processing such as enhancement processing is performed on the sound signal in the direction estimated in the signal generation processing. Since each signal processing is performed by digital signal processing, the input signal is collected by a microphone and then converted from an analog signal to a digital signal by an A / D (analog / digital) converter. A digital signal is converted into an analog signal by a D / A (digital / analog) converter.

    In the signal analysis processing unit, the input signal to both ears is collected by a microphone, the frequency analyzer performs band division for each frequency band, and the high frequency signal is further analyzed by the sound source direction estimation unit for signal estimation for direction estimation Processing is performed.

    FIG. 2 is a diagram showing a configuration of sound source direction estimation processing based on Patent Document 4. Among the signals divided for each frequency band, the level difference (ILD) and the phase difference (IPD) between the left and right signals are calculated with respect to the high frequency sound signal. Since this processing is performed in the frequency domain, a fast Fourier transform (FFT) that converts a time domain signal into the frequency domain is applied. These ILD and IPD change depending on the direction of arrival of the sound signal, and are important as clues for perception of the auditory direction. The ILD and IPD strongly depend on the path difference from the sound source to the left and right ears, that is, the body shape such as the size of the listener's head. Therefore, ILD and IPD are measured for each user for several sound source directions in advance and a database is prepared. The sound source direction estimation unit estimates the sound source direction by comparison with a database.

    On the other hand, a low-frequency signal of the band-divided signal is directly input to the signal generation processing unit based on experimentally obtained knowledge about the band that does not affect the perception of the sound source direction.

    FIG. 3 is a diagram illustrating an outline of the signal generation processing unit. The sound signal is processed for each band, and desired sound processing including filter processing is applied. For example, a sound signal enhancement process may be performed by suppressing a noise signal outside the direction of arrival. In addition, regarding the accuracy of the filter used at that time, a high-precision filter is used in a high frequency range that affects the direction perception, and a low-precision filter is allowed in a low frequency range that does not affect the direction detection.

    In general, the accuracy of a digital filter can be controlled by the filter coefficient length, but the minimum filter coefficient length necessary for the desired accuracy differs depending on the frequency band. FIG. 4 is a diagram illustrating the concept of the relationship between the filter coefficient length and the frequency band. In order to realize a filter having the same accuracy regardless of the band, the filter coefficient length becomes longer as the frequency is lower. Therefore, it means that the calculation amount of low-frequency processing increases. The filter coefficient length is an amount corresponding to the group delay, and the longest factor affects the delay of the entire system.

    If a high-precision filter is not required at low frequencies, the processing time and delay of the entire system is reduced. In the present invention, since it is an object to realize accuracy that does not affect the direction perception, the frequency that becomes the boundary between the low range and the high range is determined by the direction localization experiment.

    In the direction localization experiment, a sound signal that simulates a virtual sound source was created by digital signal processing and presented by headphones in order to simulate a state in which only a part of the band of direction perception cues deteriorates. In the simulation of the virtual sound source, the propagation characteristic from the sound source to the listener's eardrum surface is expressed as a transfer function. In particular, a transfer function in an anechoic space is called a head-related transfer function, and a method of controlling the direction of a virtual sound source by convolving a digital filter having the frequency characteristic with a sound source signal is widely performed.

    FIG. 5 is a diagram illustrating frequency amplitude characteristics of the transfer function from the sound source in the front (0 degree) direction to the left ear as an example of the head-related transfer function. In the experiment, sounds in which head-related transfer functions were convoluted with speech signals were created on a computer in advance, and the sounds were presented at random and the perceived direction of the virtual sound source was answered.

    In addition, in order to degrade the clues of perception of the low frequency direction, the band-limited interference sound was presented simultaneously by convolving the head-related transfer function and applying the low frequency filter. That is, a sound source from two directions is virtually reproduced, one is a target sound and the other is a disturbing sound, and the subject is made to answer the direction of the target sound. Experiments were performed under six conditions of 500, 1000, 1500, 2000, 20000 (full band) Hz and no interference sound for the cutoff frequency of the low-pass filter applied to limit the band of the disturbing sound.

    FIG. 6 shows the perceived direction with respect to the direction of the virtual sound source in a subject, and if they match, the answer is plotted on a diagonal line. The value of fc indicates the band of the disturbing sound. Therefore, since the answer is plotted at a position deviated from the diagonal line at 2000 Hz or higher, it can be said that the influence of the disturbing sound appears. Conversely, it can be seen that the band below 2000 Hz has little effect on the direction perception.

    From the experimental results, it is appropriate that the boundary between the low band and the high band in the frequency band dividing unit of the present invention is 2000 Hz. That is, for a band below 2000 Hz, it is possible to apply a filter process with a short coefficient length in the signal generation processing unit, or to omit the filter process. Also, in the signal analysis processing unit, if the sound source direction can be sufficiently estimated from the high frequency, the processing for the low frequency can be omitted.

    If the bandpass filter is designed to comply with JIS C1513 (Class 2), it is considered that the filter coefficient length is 256th order or less in the band of 2000 Hz or more and sufficient performance is obtained. In this case, the group delay and the calculation amount are sufficiently small, and it is possible to realize a system in which the influence on the direction perception is suppressed and the processing delay is reduced.

    From the above, it is possible to reduce the physical group delay by shortening the filter coefficient length, and to reduce the amount of calculation, the processing of the entire system is reduced, and a low delay system as a whole can be realized. .

      According to the embodiment of the present invention, for example, application to a binaural hearing aid is possible. In binaural hearing aids, localization within the head occurs if the effect on direction perception is not taken into account, but if high-precision filter processing is applied over the entire band to accurately reproduce the direction localization clues, processing delay increases. Operation in real time becomes impossible, which is problematic in terms of practicality. In the present invention, by dividing the sound signal into bands, it is possible to apply advanced acoustic signal processing while minimizing the influence on direction perception.

      According to the present invention, since the processing delay is small and power saving can be expected by reducing the load, it can be applied to the entire acoustic signal processing industry in addition to the hearing aid. In particular, it can be applied to miniaturization of audio equipment.

Claims (4)

(1) Input sound signals generated from a plurality of sound sources from both the left and right sound receiving units,
(2) The input left and right input signals are divided for each frequency band,
(3) means for estimating a sound source direction from information for each frequency band;
Based on the experimental fact that a signal in a specific frequency band does not affect the perception of the sound source direction, means for reducing signal processing in the frequency band that does not affect the perception of the sound source direction among the divided frequency bands.
(1) Means for inputting acoustic signals generated from a plurality of sound sources from both the left and right sound receiving units,
(2) means for dividing the input left and right input signals for each frequency band;
(3) means for estimating a sound source direction from information for each frequency band;
In the system for estimating the sound source direction consisting of
Based on the experimental fact that signals in a specific frequency band do not affect the perception of the sound source direction, this system reduces signal processing in the frequency band that does not affect the perception of sound source direction among the divided frequency bands. (1) Input sound signals generated from a plurality of sound sources from both the left and right sound receiving units,
(2) The input left and right input signals are divided for each frequency band,
(3) Estimate the sound source direction from the information for each frequency band,
(4) means for separating a sound source by extracting an acoustic signal in a specific sound source direction based on the sound source direction information estimated from the above;
The means for reducing the signal processing of the frequency band that does not affect the perception of the sound source direction among the divided frequency bands based on the experimental fact that the signal of the specific frequency band does not affect the perception of the sound source direction. (1) Means for inputting acoustic signals generated from a plurality of sound sources from both the left and right sound receiving units,
(2) means for dividing the input left and right input signals for each frequency band;
(3) means for estimating a sound source direction from information for each frequency band;
(4) means for separating a sound source by extracting an acoustic signal in a specific sound source direction based on the sound source direction information estimated from the above;
In a system that separates sound sources consisting of
Based on the experimental fact that signals in a specific frequency band do not affect the perception of the sound source direction, this system reduces signal processing in the frequency band that does not affect the perception of sound source direction among the divided frequency bands.