KR101296765B1 - Method and apparatus for active audio matrix decoding based on the position of speaker and listener - Google Patents

Abstract

The present invention extracts the characteristics of each speaker signal and each speaker angle from an arbitrary signal reproduced from the multi-channel speakers, decodes the stereo signal into a multi-channel signal, and extracts the signal characteristic values extracted for each channel. Compensating the decoded multi-channel signals based on the extracted signal, the power vector of each channel signal is extracted by reflecting the speaker angle of each channel in the corrected size of each channel signal, and the power vector of each channel signal Extracting the vector of the virtual sound source existing between each channel by combining, extracting the vector value of the dominant sound image by combining the extracted virtual sound sources, and extracting the vector value of each channel speaker based on the vector value of the dominant sound image The process of normalizing the position, comparing the signal magnitude of the entire decoded channel with the signal magnitude of each channel, It comprises the step of distributing the gain on each channel speaker position.

Description

Method and apparatus for active audio matrix decoding based on speaker and listener position

Figure 1 shows a conventional matrix decoder.

2 is a block diagram showing an active audio matrix decoding apparatus according to the present invention.

FIG. 3A is an embodiment of the speaker component extractor of FIG. 2.

Figure 3b is a diagram showing the redistribution of energy according to the position of the speaker and the virtual sound source for each channel according to the present invention.

4 is an embodiment of the passive matrix decoder unit of FIG.

5 is an embodiment of the channel power vector extracting unit of FIG.

FIG. 6 shows an embodiment of the virtual sound source power vector estimation unit of FIG. 2. FIG.

FIG. 7 shows an embodiment of the global power vector extracting unit of FIG.

FIG. 8 is an embodiment of the channel selector of FIG. 2.

FIG. 9 is an embodiment of the channel power distribution unit of FIG. 2.

10 is a flowchart showing an audio matrix decoding method according to the present invention.

The present invention relates to an audio reproduction system, and more particularly, to an active audio matrix decoding method and apparatus reflecting speaker and listener positions.

In general, when enjoying movies at home, terrestrial broadcasting by television broadcasting is the mainstream, but in recent years, sound of movies and the like can be heard originally by the spread of video tapes, video disks, satellite broadcasts and the like. Video tapes, video disks, satellite broadcasts, etc., which can be enjoyed with such original sound, perform matrix processing on a plurality of channels of audio signals and encode them into two-channel audio signals. Also, the two-channel audio signal encoded using the matrix processing can be reproduced in stereo. Also, when a dedicated decoder is used, five-channel audio signals of the front left (L), center (C), front light (R), left surround (Ls) and light surround (Rs) are restored from the audio signals of two channels. The center channel signal of the five-channel audio signal plays a role of obtaining a sense of stereotactic sound, and the surround channel signal enhances the clinical feeling by moving sound, environmental sound, and reverberation sound.

A commonly used matrix decoder generates the signals of the center channel and the surround channel using the sums and differences of the signals of both channels. An audio matrix in which the matrix characteristics are unchanged is known as a passive matrix decoder. Each channel signal separated by the passive matrix decoder is linearly combined by scaling down audio signals of other channels along with the corresponding channel audio signal at the time of encoding. Therefore, the signals of the channels output through the conventional passive matrix decoder have a low degree of separation between the channels, so that the sound localization is not clearly realized in the multi-channel environment. An active matrix decoder adaptively changes the matrix characteristics to improve separation of the two-channel matrix-coded code signals.

A technique related to such a matrix decoder is disclosed in US 4,799, 260 (filed 6 Feb. 1986 entitled VARIABLE MATRIX DECODER), WO 02/19768 A2 (filed 31 August 2000 entitled METHOD FOR APPARATUS FOR AUDIO MATRIX DECODING).

In a conventional matrix decoder, the gain function units 110 and 116 mainly clips the input signal to balance the levels of the stereo signals Rt and Lt. The passive matrix function unit 120 outputs a passive matrix signal from the stereo signals (R't, L't) output from the gain functions 110 and 116. The variable gain signal generator 130 generates a passive matrix. The six control signals gL. GR, gF, gB, gLB, and gRB are generated in response to the passive matrix signal generated by the function unit 120. The matrix coefficient generator 132 is a variable gain signal generator 130. 12 matrix coefficients are generated in response to the six control signals generated in the N. The adaptive matrix function unit 114 generates the input stereo signals R't and L't and the matrix coefficient generator 132. The output signals L, C, R, L, Ls, and Rs are generated in response to the matrix coefficients of the variable gain signal generator 130. The variable gain signal generator 130 monitors the level of the signal for each channel, and The optimum linear coefficient value is calculated according to the level to reconstruct the multi-channel audio signal. The castle. Matrix coefficient generation unit 132 increases the level of the channel having the highest level in a nonlinear fashion.

However, since the conventional matrix decoding system as shown in FIG. 1 does not consider the position of a virtual sound source generated in a multi-channel environment, there is a problem in that the sound image phase is not accurately positioned in space, It is difficult to accurately express it, and there is a disadvantage in that the dynamic expression ability of the image is insufficient.

The present invention is directed to an active audio matrix decoding method and apparatus for optimally tuning audio signal levels of respective channels by reflecting speaker and listener positions.

In order to solve the above technical problem, the present invention provides an audio matrix decoding method,

Extracting characteristics of each speaker signal and each speaker angle from an arbitrary signal reproduced from the multi-channel speakers;

Decoding a stereo signal into a multi-channel signal and correcting the decoded multi-channel signals based on signal characteristic values extracted for each channel;

The power vector of each channel signal is extracted by reflecting the speaker angle of each channel to the corrected magnitude of each channel signal, and the power vector of each channel signal is combined with each other to obtain a vector of a virtual sound source existing between each channel. Extraction process;

Extracting a vector value of the dominant sound image by combining the vectors of the virtual sound sources extracted in the process, and normalizing the position of each channel speaker based on the vector value of the dominant sound image;

And comparing a signal size of the entire decoded channel with a signal size of each channel and distributing a gain value to the speaker positions of the respective normalized channels.

In order to solve the above other technical problem, the present invention provides an audio matrix decoding apparatus,

A speaker component extracting unit for extracting characteristics of each speaker signal and each speaker angle from an arbitrary signal reproduced from the multi-channel speakers;

A passive matrix decoder for decoding a stereo signal into a multi-channel signal;

A signal correcting unit correcting signals of the multi-channel decoded by the decoder based on the signal characteristic value extracted by the speaker component extracting unit;

The virtual sound source power vector weight extracting the vector of the virtual sound source existing between each channel by combining the power vector of each channel signal reflecting the speaker angle of each channel to the magnitude of each channel signal corrected by the signal correction unit. government;

A global vector extractor configured to linearly combine the virtual sound source vectors estimated by the virtual sound source power vector estimator to extract a global vector representing a position and a size of a dominant sound image;

A channel selector for normalizing the position of each channel speaker based on the dominant sound image position estimated by the global vector extractor;

And a channel power distribution unit that distributes the size of each channel signal according to the ratio of the signal of each channel decoded by the passive matrix decoder to the size of the signal of the entire channel.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.

2 is a block diagram showing an active audio matrix decoding apparatus according to the present invention.

The active audio matrix decoding apparatus of FIG. 2 includes a speaker component extractor 200, a passive matrix decoder 210, a signal corrector 214, a channel power vector extractor 220, and a virtual sound source power vector estimator 230. , A global power vector extractor 240, a channel selector 250, and a channel power distributor 260.

First, a signal supply device (not shown) obtains signals from video tapes, video disks, satellite broadcasts, and the like to reproduce video signals and audio signals. In this case, the audio signal is a stereo signal of two channels that are matrix-encoded. Finally, the video signal is supplied to a monitor (not shown).

The speaker component extractor 200 extracts the characteristics of each speaker signal and each speaker angle by applying an appropriate signal processing and beamforming technique to any signal reproduced from the multi-channel speakers. That is, the speaker component extracting unit 200 extracts a gain value from the level of an arbitrary signal reproduced by the speaker, and extracts a delay from the point of time output from the speaker to the point of time input to the microphone as the delay value of the corresponding speaker. Then, the path difference of the signal received from the pair of microphones is detected and the angle of the corresponding speaker is extracted.

The passive matrix decoder 210 converts the matrix-encoded stereo signals Lt and Rt into a left channel signal L_p, a center channel signal C_p, a right channel signal R_p, (SL_p), and the light surround channel signal SR_p.

The signal corrector 214 generates the corrected signal by applying the speaker gain and delay value of each channel extracted by the speaker component extractor 200 to the signals of each channel decoded by the passive matrix decoder 210. . Accordingly, the signal corrector 214 restores the sound image of the variable speaker position to the intended sound image position. In another embodiment, the signal corrector 214 may use the speaker gain and delay value preset by the user instead of the speaker gain and delay value extracted by the speaker component extractor 200.

The channel power vector extractor 220 calculates each speaker angle extracted from the speaker component extractor 200 in addition to the magnitudes of the channel signals L_p, C_p, R_p, SL_p, SR_p corrected by the signal corrector 214. The multi-channel power vector P {L_p}, P {C_p}, P {R_p}, P {SL_p}, and P {SR_p} are extracted by multiplying. Instead of the speaker angle of each channel extracted by the extractor 200, a speaker angle value preset by the user may be used.

The virtual sound source power vector estimator 230 may extract the power vectors P {L_p}, P {C_p}, P {R_p}, P {SL_p}, and P {of each channel extracted by the channel power vector extractor 220. From SR_p}, the virtual sound source vectors vs1, vs2, vs3, vs4 and vs5 existing between each channel are calculated.

The global vector extractor 240 extracts the global power vector Gv by linear combination of the virtual sound source vectors vs1, vs2, vs3, vs4, vs5 calculated by the virtual sound source power vector estimator 230. Identify the location and size of the most dominant sound image in the sound image.

The channel selector 250 normalizes speaker positions of each channel based on the dominant sound image position corresponding to the global power vector Gv extracted by the global vector extractor 240. That is, the channel selector 250 selects channels necessary to improve the gain of the signal.

The channel power distribution unit 260 may be configured to obtain the magnitude (L_p, C_p, R_p, SL_p, SR_p) of the signal of each channel decoded by the passive matrix decoder 210 and the magnitude of the entire channel signal (Lp ² + R_p ² + C_p ^2). + SL_p ² + SR_p ² ) is compared to adjust the signal gain for each channel, and the channel selector 250 redistributes the adjusted signal gain to the position of each channel normalized. Therefore, the channel power distribution unit 260 outputs the signals L_e, R_e, C_e, SL_e, SR_e whose gains are redistributed for each channel.

3A is a detailed view of the speaker component extractor 200 of FIG. 2.

The signal generator 310 generates a digital wideband signal to generate a test signal. For example, the test signal may be a predetermined signal such as a white noise signal or an impulse noise signal.

The speaker 320 reproduces the signal generated by the signal generator 310 as sound.

The pair of microphones 330a and 330b convert sound reproduced from the speaker 320 into an electrical signal.

The signal analyzer 340 analyzes characteristics of signals input from the pair of microphones 330a and 330b.

The controller 350 extracts a gain value with the level of the signal analyzed by the signal analyzer 340, and receives the signal when the signal generator 310 generates a random signal and at a pair of microphones 330a and 330b. The signal delay value between the signals is extracted, and the speaker angle is detected by detecting a path difference between the signals received from the pair of microphones 330a and 330b through the speaker 320.

Figure 3b is a diagram showing the redistribution of energy according to the position of the speaker and the virtual sound source for each channel according to the present invention.

3B illustrates a virtual position of the speakers after the signal of each channel is corrected by the gain and delay in the signal corrector 214. That is, the layout of the reproduction speakers L, C, R, SL, and SR (shown as hatched portions) may vary depending on the position of the listener. Accordingly, the signal corrector 214 restores the distorted sound image due to the positional change of the listener to the intended original sound image through signal correction. Referring to FIG. 3B, the virtual positions L ', C', R ', SL', and SR 'of the speakers are arranged around the listener's position according to gain and delay correction.

In addition, the angles of the speakers (L, C, R, SL, and SR) of the left, center, light, left surround, and light surround channels are expressed as AngL, AngC, AngR, AngSL, and AngSR, respectively. Also, virtual sound source vectors (vs1, vs2, vs3, vs4, vs5) exist between the respective channel speakers. Also, the global power vector Gv indicates the position of the most dominant sound image among the entire sound images. Therefore, the signal level adjusted by the gain control function is redistributed to the position of the speaker of each channel normalized based on the global power vector Gv.

4 is an embodiment of the passive matrix decoder unit 210 of FIG.

The matrix encoded stereo signals Lt and Rt are fed to the left, center, right, and left surround by linear combination using the multipliers 412, 414, 422, 424, 432, 430 and the adders 410, 420, , And the audio signals (L_p, C_p, R_p, SL_p, SR_p) of the five channels of the light surround. For example, L_p = Lt, R_p = Rt, C_p = 0.7 * (Lt + Rt), SL_p = -0.866Lt + 0.5Rt, SR_p = -0.5Lt + 0.866Rt.

FIG. 5 is an embodiment of the channel power vector extractor 220 of FIG. 2.

Referring to FIG. 5, the first, second, third, fourth, and fifth squarers 512, 514, 516, 518, and 519 are left, center, and right decoded by the passive matrix decoder 210. Each power value is calculated by squaring the signals L_p, C_p, R_p, SL_p and SR_p of the left surround and right surround channels.

The first multiplier 532 multiplies the power value of the left channel signal calculated by the first multiplier 512 by the angle AngL of the left channel speaker (for example, 120 degrees) to extract the power vector of the left channel ( P {L_p}) is extracted.

The second multiplier 534 multiplies the power value of the light channel signal calculated by the second multiplier 514 by the angle AngR (for example, 60 degrees) of the extracted light channel speaker to obtain the power vector of the light channel ( P {R_p}) is extracted.

The third multiplier 536 multiplies the power value of the center channel signal calculated by the third multiplier 516 by the angle AngC of the center channel speaker (for example, 90 degrees) to obtain the power vector P of the center channel. {C_p}).

The fourth multiplier 538 multiplies the power value of the right surround channel signal calculated by the fourth multiplier 518 by the angle AngL of the left surround channel speaker (for example, 200 degrees) to extract the power of the left surround channel. The vector P {SL_p} is extracted.

The fifth multiplier 539 multiplies the power value of the left surround channel signal calculated by the fifth squarer 519 by the angle AngSR (for example, 340 degrees) of the light surround channel speaker extracted, and thus the power of the light surround channel. The vector P {SR_p} is extracted.

FIG. 6 is an embodiment of the virtual sound source power vector estimator 230 of FIG. 2.

The first adder 610 extracts the first virtual sound source vector value (vs1) by adding the left channel power vector P {L_p} and the center channel power vector P {C_p}.

The second adder 620 adds the power vector P {C_p} of the center channel and the power vector P {R_p} of the write channel to extract a second virtual sound source vector value (vs2).

The third adder 630 adds the power vector P {R_p} of the write channel and the power vector P {SR_p} of the light surround channel and extracts the third virtual sound source vector value (vs3).

The fourth adder 640 extracts the fourth virtual sound source vector value vs4 by adding the power vector P {SR_p} of the right surround channel and the power vector P {SL_p} of the left surround channel.

The fifth adder 650 adds the power vector P {SL_p} of the left surround channel and the power vector P {L_p} of the left channel to extract the fifth virtual sound source vector value (vs5).

FIG. 7 is an embodiment of the global power vector extractor 240 of FIG. 2.

The first, second, third, fourth and fifth virtual sound source vector values (vs1, vs2, vs3, vs4, vs5) are linearly combined through adders 710, 720 and 730, . This global vector value Gv represents the position and size of the most dominant sound image among all the sound images as shown in FIG.

FIG. 8 is an embodiment of the channel selector 250 of FIG. 2.

The first subtractor 826 subtracts the position value of the global vector value Gv from the angle AngR of the left channel speaker to obtain the speaker position θ _ch1 of the normalized left channel.

The second subtractor 824 subtracts the position value of the global vector value Gv from the angle AngL of the light channel speaker to obtain the position θ _ch2 of the normalized light channel speaker.

The third subtractor 822 subtracts the position value of the global vector value Gv from the angle AngC of the center channel speaker to obtain the normalized center channel speaker position θ _ch3 .

The fourth subtractor 818 calculates the position θ _ch4 of the normalized left surround channel speaker by subtracting the position value of the global vector value Gv from the angle AngRS of the left surround channel speaker.

The fifth subtractor 816 calculates the position θ _ch5 of the normalized light channel speaker by subtracting the position value of the global vector value Gv from the angle AngSL of the light surround channel speaker.

FIG. 9 is an embodiment of the channel power distribution unit 260 of FIG. 2.

The first, second, third, fourth, and fifth multipliers 922, 924, 926, 928, and 929 are normalized channel position values θ _ch1 , θ _ch2 , θ _ch3 , θ _ch4 , and θ _ch5. ) As a parameter gain control function f (x) (912, 914, 916, 918, 919) and the signal magnitude values L_p, R_p, C_p, SL_p, SR_p of the decoded channel as parameters. The signals ((L_e, R_e, C_e, SL_e, SR_e) of each redistributed channel are output by multiplying g (x)) 922, 924, 926, 928, and 929.

At this time, the gain control function g (x) adjusts the size of each channel signal according to the ratio of the signal size of each channel to the size of the entire channel signal by comparing the size of the decoded channel- do. For example, if the signal size (R_p) of the write channel is 20% or more of the signal size (L_p ² + R_p ² + C_p ² + SL_p ² + SR_p ² ) of the entire channel, (R_p). If the signal size R_p of the write channel is 20% or less of the total signal magnitude (L_p ² + R_p ² + C_p ² + SL_p ² + SR_p ² ), the magnitude R_p of the write channel is reduced in proportion to the logarithm function .

10 is a flowchart showing an audio matrix decoding method according to the present invention.

First, the gain and delay of each speaker signal and the angle of each speaker are extracted from an arbitrary signal reproduced from the speakers of the multichannel (step 1008).

Subsequently, the matrix encoded stereo signal is decoded into a multi-channel signal through a predetermined passive matrix decoding algorithm (step 1010).

Subsequently, each channel speaker gain and delay value is applied to the signals of each decoded channel to generate a corrected signal.

Subsequently, the power vector of the signal for each channel is calculated by multiplying the corrected magnitude of each channel signal by the angles of the extracted channels.

Subsequently, the power vector values of the respective channels are linearly combined with each other to extract a vector of the virtual sound source existing between the respective channels (step 1030).

Subsequently, a global vector indicating the position of the dominant sound image is calculated by linear combination of the extracted virtual sound sources, and the position of each channel speaker is normalized based on the position of the dominant sound image (step 1050).

Subsequently, by comparing the size of the decoded channel total signal and the signal size of each channel, the size of each channel signal is adjusted according to the ratio of the signal size of each channel to the size of the channel total signal, and the adjusted signal size for each channel ( Energy) is redistributed to the normalized speaker positions (step 1060).

The present invention is not limited to the above-described embodiment, and of course, modifications may be made by those skilled in the art within the spirit of the present invention. In another embodiment, the present invention can extend a 5-channel sound source to a 7-channel sound source. That is, the seven-channel matrix decoder includes a channel expansion unit (not shown) and a channel power enhancement unit (not shown). The channel extension unit creates a sound source corresponding to the left back channel and the right back channel from the sound sources of five channels using a vector project method, and newly generates the power of the surround left channel signal and the surround right channel signal. Readjust the effect of the back channel and right back channel signals.

For example, the channel extension unit derives the values of the left back channel and the right back channel at positions 40 and 100 from the left and right surround channels (SL, SR) of five channels, respectively, using the vector project equation. do.

In addition, the channel power enhancement unit selects the degree to be improved for each channel according to the position of the entire sound image and redistributes the power for each channel according to the layout using a separate nonlinear function. In other words, the channel power enhancement unit selects the degree of enhancement for each channel according to the position of the entire sound image of the seven channels. Therefore, the channel power improvement unit determines the degree to be improved according to each channel position by a function, and redistributes the power of the entire channel by recalculating the power for each channel using a separate nonlinear function g (x).

The present invention can also be embodied as computer readable codes on a computer readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, hard disk, floppy disk, flash memory, optical data storage, And the like. The computer readable recording medium may also be distributed over a networked computer system and stored and executed as computer readable code in a distributed manner.

As described above, according to the present invention, each channel signal level can be optimally tuned based on the position of the virtual sound image generated in consideration of the speaker and listener positions. Therefore, it is possible to psychologically solve the problem of low separation due to the limitation of the existing matrix decoder, that is, the high correction inevitably generated between channels. In addition, it is possible to restore the distorted sound image due to the change of the position of the listener to the intended original sound image through signal correction.

Claims (15)

A method for decoding an audio matrix, Extracting characteristics of each speaker signal and each speaker angle from an arbitrary signal reproduced from the multi-channel speakers; Decoding a stereo signal into a multi-channel signal and correcting the decoded multi-channel signals based on signal characteristic values extracted for each channel; The power vector of each channel signal is extracted by reflecting the speaker angle of each channel in the corrected magnitude of each channel signal, and the vector of virtual sound sources existing between each channel is extracted based on the power vector of the signal of each channel. Process of doing; Extracting a vector value of the dominant sound image by combining the vectors of the virtual sound sources extracted in the above process, and normalizing the position of each channel speaker based on the vector value of the dominant sound image; Comparing the signal magnitude of the entire decoded channel with the signal magnitude of each channel and distributing a gain value to the speaker positions of the respective normalized channels; Correcting the signals of the multi channel And generating gain-delay-corrected signals by applying the signal gain and delay value extracted for each channel to the signals of each decoded channel. The method of claim 1, wherein the feature extraction process of the speaker signal is performed. The gain value is extracted from the level of any signal played by the speaker, And extracting a signal delay value from a time point output from the corresponding speaker to a time point input to the microphone. The method of claim 1, wherein the speaker angle extraction process An audio matrix decoding method comprising detecting a path difference of a signal received from a pair of microphones through a speaker and extracting an angle of the corresponding speaker. delete The method of claim 1, wherein the signal correction process is performed. And generating gain-delay corrected signals by applying a signal gain and a delay value preset by a user to the signals of each of the decoded channels. The method of claim 1, wherein the power vector extraction process is performed. Calculating a power value by squaring the signals of the decoded ga channel; And calculating a power vector of a signal for each channel by multiplying the speaker angle for each channel and the power value. The method of claim 1, wherein the virtual sound source vector extraction process And summing the power vector value of the specific channel and the power vector value of the channel adjacent to the channel. The method of claim 1, wherein the normalized position value calculation process is performed. Calculating a vector of the dominant sound image by linearly combining the extracted vectors of the virtual sound sources; And calculating the normalized position value of the speaker for each channel by subtracting the dominant sound image from the angle of the speaker for each channel. The method of claim 1, wherein the power distribution process Adjusting a size of each channel signal according to a ratio of a signal size of each channel to a size of the entire channel signal by comparing the size of the decoded channel whole signal with a signal size of each channel; And multiplying the signal size adjusted for each channel and the position value of each normalized channel. An audio matrix decoding apparatus comprising: A speaker component extracting unit for extracting characteristics of each speaker signal and each speaker angle from an arbitrary signal reproduced from the multi-channel speakers; A passive matrix decoder for decoding a stereo signal into a multi-channel signal; A signal correcting unit correcting signals of the multi-channel decoded by the decoder based on the signal characteristic value extracted by the speaker component extracting unit; The virtual sound source power vector weight extracting the vector of the virtual sound source existing between each channel by combining the power vector of each channel signal reflecting the speaker angle of each channel to the magnitude of each channel signal corrected by the signal correction unit. government; A global vector extractor configured to linearly combine the virtual sound source vectors estimated by the virtual sound source power vector estimator to extract a global vector representing a position and a size of a dominant sound image; A channel selector for normalizing the position of each channel speaker based on the dominant sound image position estimated by the global vector extractor; A channel power distribution unit for distributing the size of each channel signal according to a ratio of the size of the signal of each channel decoded by the passive matrix decoder to the size of the signal of all channels; The signal correction unit And applying the gain and the delay value extracted for each channel to the signals of each decoded channel to generate the gain and delay corrected signals. The method of claim 10, wherein the speaker component extraction unit A signal generator for generating an arbitrary signal; A speaker unit for reproducing a signal generated by the signal generator; A pair of microphone units for converting sound reproduced from the speaker unit into electrical signals; Extracts a gain value with a level of a signal input from the microphone unit, extracts a signal delay value between a time point at which the signal generator generates an arbitrary signal and a signal output from the microphone unit, and pairs an upper pair through the speaker unit And a controller for detecting a path difference of a signal received from a microphone unit and extracting an angle of the speaker unit. The apparatus of claim 10, wherein the virtual sound source power vector estimator comprises: a squarer configured to calculate respective power values by squaring signals of the decoded multi-channel; A multiplier for extracting a power vector of each channel by multiplying the loudspeaker angle by the magnitudes of the channel signals calculated by the squarer; And an adder for adding a vector value of a specific channel signal extracted by the multiplier and a vector of channels adjacent to the channel. The method of claim 10, wherein the channel selector And a subtractor which subtracts the position of the dominant sound image extracted by the global vector extractor at a specific channel speaker angle. The method of claim 10, wherein the channel power distribution unit And a multiplier for outputting signals of each redistributed channel by multiplying a batch function having the position values of the normalized channel as parameters and a gain adjustment function having the signal magnitude values of the decoded channel as parameters. Device. 15. The method of claim 14, wherein the gain control function increases the size of a specific channel signal when the size of the signal of the decoded specific channel is greater than or equal to a certain ratio to the size of the signal of the entire channel, and the size of the signal of the decoded specific channel. Is less than a certain ratio to the magnitude of the signal of the entire channel, the audio matrix decoding device characterized in that for reducing the magnitude of the specific channel signal.

Images