JP6006627B2 - Impulse response length conversion device, impulse response length conversion method, impulse method conversion program - Google Patents

Description

  The present invention relates to an apparatus, a method, and a program for performing impulse response length conversion, and more particularly, to express a temporal expression of a head related transfer function in order to calculate a three-dimensional acoustic signal in a device such as a mobile phone that does not have high calculation capability. The present invention relates to a technique for shortening the filter length of the impulse response, while maintaining the localization quality.

  Multi-channel sound systems such as the 5.1ch surround system and the 22.2 multi-channel system for Super Hi-Vision have a feature that a sound source can be localized in various directions in a three-dimensional space (for example, non-patent literature). 1). Here, using a head-related transfer function (HRTF) measured in advance, the acoustic space of the three-dimensional acoustic signal can be reproduced in a pseudo manner using headphones. In this case, an impulse response that is a time expression of the measured head-related transfer function is folded on the sound source as a directional filter. In general, the length of a binaural impulse response, which is a time expression of the head-related transfer function, is determined by a measurement method.

Akio Ando “Highly Realistic Acoustic Technology and Its Theory” IEICE Fundamental Review Vol.3 No.4 pp.33-46 Apr 2010

  3D audio signals are produced using many audio signal channels, as represented by, for example, 22.2ch audio for Super Hi-Vision, so it is expensive to perform convolution processing on multiple channels simultaneously. Ability is required. When a device such as a portable terminal that does not have high calculation capability is used, in order to implement three-dimensional sound, it is desirable to reduce the amount of calculation by making the filter length of the impulse response as short as possible.

  In the measurement of the head-related transfer function, the impulse response, which is the time expression of the transfer function, is measured by the swept-sine method or the method using the M-sequence, assuming that the transfer system is an LTI system (linear time-invariant system). . In the measured impulse response, reflected sound from a measuring device in the laboratory is mixed as a measurement error, and there are some which can separate reflected sound components measured with a delay on the time axis of the impulse response. However, it is very difficult to directly cut the filter length of the impulse response, which is a time expression, to a desired length while maintaining the characteristics of the head-related transfer function, particularly the characteristics related to the amplitude frequency characteristic.

  In addition, there is a method to simplify the cranial transfer function with a cepstrum for the purpose of shortening the filter length of the binaural impulse response, but it is optimal to include the interaural time difference and level difference of the binaural impulse response. It is difficult to select the filter length.

  Accordingly, an object of the present invention made in view of such a point is to provide an impulse response length conversion device, an impulse response length conversion method, an impulse, and a filter length that can be shortened using a cepstrum while having characteristics related to the amplitude frequency characteristics of the impulse response. It is to provide a method conversion program.

  In order to solve the above-described problems, an impulse response length conversion apparatus according to the present invention includes an impulse response storage unit that stores an original impulse response that is a temporal expression of a head related transfer function, and a filter length that is lower than the filter length of the original impulse response. A regression model expressing the relationship between the cepstrum order, the amplitude ratio, and the filter length based on the amplitude ratio between the maximum amplitude of the general impulse response and the other amplitudes, which are converted from the original impulse response expressed by the order cepstrum. A cepstrum order determining unit that calculates and determines a cepstrum order for a desired amplitude ratio and a desired filter length based on the regression model, and a general impulse response that generates a general impulse response of the original impulse response with the determined cepstrum order A generator and an impulse of the desired filter length using the approximate impulse response Comprising the impulse response reconstruction unit which reconstructs a response, the.

  An impulse response conversion method according to the present invention is an impulse response length conversion method in an impulse response length conversion device including an impulse response storage unit that stores an original impulse response that is a time expression of a head related transfer function, The amplitude ratio between the maximum amplitude of the general impulse response obtained by re-converting the original impulse response expressed by a cepstrum of an order lower than the filter length of the original impulse response into an impulse response and the other amplitude. A step of calculating a regression model representing a relationship between a cepstrum order, an amplitude ratio and a filter length; a step of determining a cepstrum order for a desired amplitude ratio and a desired filter length based on the regression model; and the determined cepstrum order To generate a rough impulse response of the original impulse response. Comprising the steps of, and reconstructing an impulse response of the desired filter length using the outline impulse response, the.

  Further, an impulse response conversion program according to the present invention includes an impulse response length conversion device including an impulse response storage unit that stores an original impulse response that is a time expression of a head related transfer function, and an order lower than the filter length of the original impulse response. The regression model expressing the relationship between the cepstrum order, the amplitude ratio, and the filter length is calculated from the amplitude ratio between the maximum amplitude of the general impulse response, which is the original impulse response expressed in the cepstrum of the above, and converted into an impulse response. Determining a cepstrum order for a desired amplitude ratio and a desired filter length based on the regression model, generating a rough impulse response of the original impulse response with the determined cepstrum order, and the rough shape Impulse of the desired filter length using an impulse response And reconstructing an answer, to the execution.

  According to the impulse response length converter, the impulse response length conversion method, and the impulse method conversion program according to the present invention, it is possible to shorten the filter length using the cepstrum while maintaining the characteristics related to the amplitude frequency characteristics of the impulse response. .

It is a figure which shows the structure of the impulse response length converter which concerns on one Embodiment of this invention. It is a figure which shows an example of the amplitude ratio of each sample in a rough impulse response. It is a figure which shows an example of the regression model showing the relationship between cepstrum order, filter length, and amplitude ratio.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a diagram showing a configuration of an impulse response length conversion apparatus according to an embodiment of the present invention. The impulse response length conversion apparatus 1 includes an impulse response storage unit 10, a cepstrum order determination unit 20, a rough impulse response generation unit 30, an impulse response division unit 40, an all-pass component time difference acquisition unit 50, and an impulse response reconfiguration. Unit 60.

  The impulse response storage unit 10 is a database of head-related transfer functions, and stores original impulse responses, which are temporal expressions of head-related transfer functions, for both ears of a plurality of persons for each three-dimensional space direction. For example, the filter length (impulse response length) of the original impulse response is composed of 512 samples in the time direction, and the original impulse responses of the right ear and the left ear are stored as a pair.

  The cepstrum order determination unit 20 determines the order of the cepstrum that models the original impulse response in order to shorten the filter length while maintaining the characteristics regarding the amplitude frequency characteristics and localization quality of the original impulse response stored in the impulse response storage unit 10. To do. Specifically, first, the cepstrum order determination unit 20 obtains a rough impulse response obtained by reconverting the original impulse response expressed by a cepstrum having an order lower than the filter length of the original impulse response into an impulse response, and the rough impulse response. Based on the amplitude ratio between the maximum amplitude and other amplitudes, a regression model representing the relationship between the cepstrum order, the amplitude ratio and the filter length is calculated. Hereinafter, calculation of the regression model will be described in detail.

  The cepstrum order determination unit 20 extracts a pair of original impulse responses of both ears from the impulse response storage unit 10 and calculates a real cepstrum by equation (1) for each original impulse response of one ear. Where imp (n) is the original impulse response of one ear for calculation, abs () is the absolute value, FFT () is the discrete Fourier transform, IFFT is the discrete Fourier inverse transform, and Re () is the process of extracting the real number of the complex number , Log () represents natural logarithm processing, respectively.

  The zeroth-order coefficient of the real cepstrum calculated by Equation (1) is a bias term and expresses the magnitude of the entire original impulse response. Further, the frequency envelope of the original impulse response can be expressed only by the low-order coefficient of the real cepstrum. For this reason, the cepstrum order determination unit 20 converts the cepstrum expressed by the lower order with the zeroth order and higher order coefficients of the real cepstrum as 0, and converts the cepstrum to the impulse response again. The amplitude ratio with the amplitude is obtained.

  For example, the cepstrum order determination unit 20 sets the 0th order and the coefficients from the 51st order to the 511th order to 0 for the real number cepstrum calculated by Expression (1). At this time, the cepstrum order determination unit 20 performs a process of doubling the coefficient with respect to the coefficient of the order (for example, the first order to the 50th order) for which the coefficient is not 0 in order to save energy.

  The cepstrum order determination unit 20 converts the cepstrum expressed by a low order into an impulse response again using the equation (2). Here, exp () represents an exponential function.

  The impulse response expressed by the equation (2) is a time expression of the frequency response from which the fine structure is removed by the cepstrum expressed by a low order. Hereinafter, the impulse response after reconversion represented by Expression (2) is referred to as a general impulse response. The cepstrum order determination unit 20 obtains an amplitude ratio between the maximum amplitude of the approximate impulse response and other amplitudes. In the approximate impulse response, the maximum amplitude appears at a location close to time 0. Therefore, the cepstrum order determination unit 20 uses the amplitude value of the approximate impulse response at time 0 as a reference to determine the amplitude ratio d [dB] of each sample (n). ] Is calculated by equation (3).

  FIG. 2 is a diagram illustrating an example of the amplitude ratio of each sample in the rough impulse response. FIG. 2 shows the amplitude ratio represented by Equation (3) for a given cepstrum order (for example, 50th order), and the cepstrum order determination unit 20 stores the impulse response storage unit 10 while changing the cepstrum order. Equations (1) to (3) are calculated for all the stored original impulse responses, and the amplitude ratio of each sample in the approximate impulse response is calculated.

  The cepstrum order determination unit 20 calculates a regression model representing the relationship between the cepstrum order, the amplitude ratio, and the filter length from the calculations of the equations (1) to (3) for all the original impulse responses. FIG. 3 is a diagram illustrating an example of a regression model representing the relationship between the cepstrum order, the filter length, and the amplitude ratio. In the case of FIG. 3, the cepstrum order determination unit 20 changes the cepstrum order from the 30th order to the 80th order at the 10th order interval, calculates the equations (1) to (3) for all the original impulse responses, and obtains the cepstrum. A linear model representing the relationship between the order, the filter length, and the amplitude ratio is calculated. Note that the cepstrum order determination unit 20 can calculate a regression model using any suitable model, not limited to a linear model.

  The cepstrum order determination unit 20 determines a cepstrum order for a desired amplitude ratio and a desired filter length based on the calculated regression model. The desired amplitude ratio indicates the amplitude ratio of a sample whose coefficient should be maintained (not 0) with respect to the maximum amplitude of the approximate impulse response. The approximate impulse response represents the frequency envelope of the original impulse response. That is, the desired amplitude ratio indicates how much the amplitude frequency characteristic and localization quality of the original impulse response should be maintained in the shortened impulse response. The desired filter length is a filter length that is shorter than the filter length of the original impulse response, and indicates a shortened filter length that is used when three-dimensional sound is mounted on a device that does not have high calculation capability such as a portable terminal. is there.

  For example, if the desired amplitude ratio and the desired filter length are 90 dB and 256 samples, respectively, the cepstrum order determination unit 20 determines the cepstrum order from the 90 dB regression model of FIG. Here, it is necessary to provide a margin before and after the filter from the time difference between the left and right time differences and the head-related transfer functions in other directions. If 50 samples before and after the 256-sample filter are used as a margin, the required filter length is calculated as 256−50 × 2 = 156. The cepstrum order determining unit 20 can determine the cepstrum order 55 when the filter length is 156 by the 90 dB regression line of FIG.

  The rough impulse response generation unit 30 holds coefficients from the 0th order to the cepstrum order (for example, 55th order) calculated by the cepstrum order determination unit 20, sets the other coefficients to 0, and uses the original impulse according to Equation (2). Generate a general impulse response imp_g of the response. The rough impulse response generator 30 outputs the generated rough impulse response imp_g to the impulse response reconstructor 60.

  The impulse response dividing unit 40 calculates the minimum phase transfer function using only the right coefficient part of the real cepstrum shown in Equation (1), and calculates the all-pass transfer function by dividing the original impulse response by the minimum phase transfer function. To do. The impulse response dividing unit 40 calculates an all-pass transfer function for a pair of binaural original impulse responses, and outputs the calculated all-pass transfer function to the all-pass component time difference acquisition unit 50. Further, the impulse response dividing unit 40 calculates the maximum peak time t_ip of the original impulse response of the reference ear (for example, the left ear) from the pair of original impulse responses of both ears, and impulses the calculated maximum peak time t_ip. The data is output to the response reconstruction unit 60.

  The all-pass component time difference acquisition unit 50 calculates the time difference t between the left and right with the cross-correlation function for each all-pass component of each pair of binaural original impulse responses. For example, if the reference ear is the left ear, when t <0, it represents that the all-pass component of the original impulse response of the right ear reaches a peak earlier than the left ear, and when t> 0, the original impulse response of the left ear. This means that the all-pass component of reaches the peak earlier than the right ear. The all-pass component time difference acquisition unit 50 outputs the left / right time difference t to the impulse response reconstruction unit 60.

  The impulse response reconstruction unit 60 reconstructs an impulse response having a desired filter length (for example, 256 samples) from the approximate impulse response imp_g, the left and right time difference t, and the maximum peak time t_ip of the impulse response. First, the impulse response reconstruction unit 60 performs a right shift based on the maximum peak time t_ip with respect to the rough impulse response imp_g of the reference ear (for example, the left ear). Here, the left margin of the impulse response is set to 50 samples, and the shift is performed in consideration of an offset value so that the maximum peak time t_ip of the left ear impulse response of the head-related transfer function having a sound source localization on the median plane is 50 samples. Find the amount.

The impulse response reconstruction unit 60 can obtain the offset amount and the binaural shift amount by the following equations.
Offset value = Maximum peak time t_ip−50 for left ear impulse response with median plane localization
Right shift amount (left ear) = left ear impulse response maximum peak time t_ip in each direction-offset value Right shift amount (right ear) = right shift amount (left ear) + t

  As described above, the impulse response reconstruction unit 60 can shift the approximate impulse response imp_g in the time direction to reconstruct the shortened impulse response for each pair of binaural original impulse responses. The impulse response reconstructed by the impulse response reconstruction unit 60 has a filter length (for example, 256 samples) shorter than the filter length (for example, 512 samples) of the original impulse response, and an approximate impulse with a cepstrum order that satisfies a desired amplitude ratio. By using the response imp_g, the amplitude frequency characteristic and localization quality of the original impulse response are also maintained.

  Thus, according to the present embodiment, the cepstrum order determination unit 20 calculates the maximum amplitude of the approximate impulse response obtained by reconverting the original impulse response expressed by the cepstrum having an order lower than the filter length of the original impulse response into the impulse response. A regression model that represents the relationship between the cepstrum order, the amplitude ratio, and the filter length is calculated from the amplitude ratio with other amplitudes, and based on the regression model, the cepstrum order for the desired amplitude ratio and the desired filter length is determined. The impulse response generation unit 30 generates a rough impulse response of the original impulse response with the determined cepstrum order, and the impulse response reconstruction unit 60 reconstructs an impulse response of a desired filter length using the rough impulse response. As a result, the filter length can be shortened while maintaining the characteristics related to the amplitude frequency characteristic of the original impulse response. That is, in order to reduce the calculation load when reproducing a three-dimensional sound signal with headphones using a device that does not have high calculation ability such as a mobile phone, the amplitude frequency characteristic of the impulse response for calculating the three-dimensional sound signal, The filter length can be optimally shortened while maintaining the localization quality.

  Although the present invention has been described based on the drawings and examples, it should be noted that those skilled in the art can easily make various modifications and corrections based on the present disclosure. Therefore, it should be noted that these variations and modifications are included in the scope of the present invention. For example, the functions included in each component, each step, etc. can be rearranged so that there is no logical contradiction, and multiple components, steps, etc. can be combined or divided into one It is.

  It should be noted that the present invention can be realized as a program for causing the processor of the impulse response length converter 1 to execute an equivalent process (step), and it is understood that these are also included in the scope of the present invention. I want. For example, the impulse response length conversion apparatus 1 can store a program describing processing contents for realizing each function in a storage unit, and read and execute the program by a central processing unit (CPU).

  According to the present invention, in order to calculate a three-dimensional acoustic signal in a device that does not have high calculation capability such as a mobile phone, the filter length of the impulse response, which is a temporal expression of the head related transfer function, is changed to the amplitude frequency characteristic, localization It becomes possible to shorten while maintaining quality.

DESCRIPTION OF SYMBOLS 1 Impulse response length conversion apparatus 10 Impulse response memory | storage part 20 Cepstrum order determination part 30 Schematic impulse response production | generation part 40 Impulse response division | segmentation part 50 All pass component time difference acquisition part 60 Impulse response reconstruction part

Claims (3)

An impulse response storage unit for storing an original impulse response which is a time expression of the head-related transfer function;
The cepstrum order, the amplitude ratio, and the filter based on the amplitude ratio between the maximum amplitude of the general impulse response expressed by the cepstrum of the order lower than the filter length of the original impulse response and converted into the impulse response and the other amplitudes. A cepstrum order determination unit that calculates a regression model representing a relationship with the length, and determines a cepstrum order for a desired amplitude ratio and a desired filter length based on the regression model;
A general impulse response generator for generating a general impulse response of the original impulse response at the determined cepstrum order;
An impulse response length conversion device comprising: an impulse response reconstruction unit that reconstructs an impulse response of the desired filter length using the approximate impulse response. An impulse response length conversion method in an impulse response length conversion device including an impulse response storage unit that stores an original impulse response that is a time expression of a head related transfer function,
The processing procedure by the impulse response length converter is as follows:
The cepstrum order, the amplitude ratio, and the filter based on the amplitude ratio between the maximum amplitude of the general impulse response expressed by the cepstrum of the order lower than the filter length of the original impulse response and converted into the impulse response and the other amplitudes. Calculating a regression model representing the relationship with length;
Determining a cepstrum order for a desired amplitude ratio and a desired filter length based on the regression model;
Generating a rough impulse response of the original impulse response with the determined cepstrum order;
Reconstructing an impulse response of the desired filter length using the approximate impulse response. In an impulse response length conversion device including an impulse response storage unit that stores an original impulse response that is a time expression of a head-related transfer function,
The cepstrum order, the amplitude ratio, and the filter based on the amplitude ratio between the maximum amplitude of the general impulse response expressed by the cepstrum of the order lower than the filter length of the original impulse response and converted into the impulse response and the other amplitudes. Calculating a regression model representing the relationship with length;
Determining a cepstrum order for a desired amplitude ratio and a desired filter length based on the regression model;
Generating a rough impulse response of the original impulse response with the determined cepstrum order;
Reconstructing an impulse response of the desired filter length using the approximate impulse response, and executing an impulse response length conversion program.

Images