JP4424348B2 - Sound image localization device - Google Patents

Description

  The present invention realizes the rear virtual sound image localization by outputting the sound of the rear channel signal processed using the head related transfer function simulating the spatial propagation characteristic from the surroundings to the human ear from the front speaker. The present invention relates to a sound image localization apparatus.

  2. Description of the Related Art Conventionally, there has been disclosed an apparatus that realizes various sound image localizations using a model head-related transfer function (hereinafter referred to as head-related transfer function) that simulates spatial propagation characteristics from the surroundings to the human ear. In addition, since it is not practical and practical to place a multi-channel speaker, the virtual sound localization in the rear is performed by performing crosstalk cancellation that cancels this spatial propagation characteristic and adding the sound localization in the rear. Has been proposed (see Patent Document 1). This crosstalk cancellation is considered to be necessary as a premise for adding rear localization. That is, it is considered necessary to add the rear localization after canceling the above-described propagation characteristics in order to realize accurate sound image localization.

  This crosstalk cancellation is a signal that reversely transforms the head-related transfer function that simulates the propagation characteristics from the front speaker, and puts the sound of the left front speaker into the left ear and the sound of the right front speaker into the right ear. It performs processing and gives the effect of listening with headphones.

In Patent Document 1, a method of canceling crosstalk as shown in FIG. 19 is illustrated.
JP 2001-86599 A

  However, the crosstalk cancellation generally requires an inverse conversion operation, and there is a problem that processing is large. Further, the manner in which the sound propagates to the ears varies depending on the individual because the manner in which the sound is diffracted differs depending on the width of the face. Depending on the individual difference, there is a case where the effect of the rearward virtual sound localization that can be heard from the rear is not achieved at all. In addition, the effect of sound image localization is sensitive to the installation angle of the speaker and the orientation of the face, and there is a problem that the effect can be achieved only in a pinpoint manner.

  Therefore, an object of the present invention is to realize a rear virtual sound image localization more reliably by a simple calculation in a sound image localization device that realizes a rear virtual sound image localization.

  In the present invention, means for solving the above-described problems are configured as follows.

(1) The present invention
LD, a head-related transfer function that simulates spatial propagation from the real speaker FL provided on the left front to the left ear.
LC, a head-related transfer function that simulates spatial propagation from the real speaker FL to the right ear,
RC, a head-related transfer function simulating spatial propagation from the real speaker FR provided on the front right to the left ear
A head-related transfer function simulating spatial propagation from the real speaker FR to the right ear is RD,
The center line of the listener's left and right direction passing through the center of the listener's head is L,
A head-related transfer function simulating spatial propagation from the virtual speaker VL to the left ear, which is set symmetrically with the real speaker FL with respect to the center line L, is RLD,
RLC, a head related transfer function simulating spatial propagation from the virtual speaker VL to the right ear,
A head-related transfer function simulating spatial propagation from the virtual speaker VR to the left ear that is set in line symmetry with the real speaker FR with respect to the straight line L is RRC,
RRD is a head related transfer function that simulates spatial propagation from the virtual speaker VR to the right ear.
An L direct output unit that outputs the audio signal of the audio input channel on the left rear by applying a filter having characteristics obtained by dividing RLD by LD;
An L cross output unit that outputs the audio signal of the audio input channel on the left rear by applying a filter having characteristics obtained by dividing RLC by LC;
An R-cross output unit that outputs the audio signal of the rear right audio input channel by applying a filter having characteristics obtained by dividing RRC by RC;
An R direct output unit that outputs the audio signal of the rear right audio input channel by applying a filter having characteristics obtained by dividing RRD by RD;
A first adder that adds the difference signal between the output signal of the L direct output unit and the output signal of the R cross output unit to the audio signal of the front left audio input channel, and outputs the difference signal;
A sound image comprising: a second addition unit that adds a difference signal between the output signal of the R direct output unit and the output signal of the L cross output unit to the audio signal of the front right audio input channel and outputs the resultant signal. It is a stereotaxic device.

  According to the L direct output unit, L cross output unit, R cross output unit, and R direct output unit of the present invention, the audio signal of the audio input channel for the rear is processed. Since the calculation for filtering the audio signal is merely applying a filter having a characteristic obtained by dividing the transfer function, the sound image localization apparatus can be configured by a simple calculation.

  Further, according to the apparatus of the present invention, the sound is output more reliably from the rear than the signal processing by the crosstalk cancellation by the inverse matrix calculation in accordance with the previous theory, even though it is output from the front speaker. It has been confirmed by the inventors' experiment that it is perceived as The reason why the apparatus of the present invention can obtain a result better than the calculation based on the theory is that the previous theory uses a model of the observation result of one head-related transfer function, and is different from the actual listener. So, it doesn't always work as expected. Therefore, even if the present invention gives better results than this theoretical calculation, it does not violate the law of nature.

  Further, according to the present invention, it is not sensitive to the orientation of the listener's face, and even if the listener moves back and forth with respect to the actual speaker installed in the front, the sense that sound is output from the rear is not impaired. This has been confirmed by the inventors' experiments. This is presumed that the present invention skillfully captures the human sense of “perceived as if sound is being output from behind” that is not easily influenced by the direction of the sound source.

  Here, to give an example of the configuration of (1), the rear localization adding unit 131 of FIG. 1 to be described later corresponds to the illustrated output unit and addition unit of the present invention. However, the present invention is not limited to this.

  In the present invention, the characteristic obtained by dividing RLD by LD is a gain characteristic obtained by dividing the gain of RLD by the gain of LD for each frequency. The same applies to the L cross output unit, the R cross output unit, and the R direct output unit.

  A real speaker refers to a speaker that is actually installed, and is a concept opposite to a virtual speaker that is not actually installed.

(2) The present invention
The filter calculation unit sets both the speaker and the virtual speaker symmetrically with respect to the front direction of the listener, and sets the model head-related transfer function to LD = RD, LC = RC, RLD = RRD, RLC = RRC.

  If comprised in this way, since a head-related transfer function can be made the same on either side, simplification can be anticipated further from the structure of (1). In addition, since the left and right head related transfer functions are exactly the same, it is possible to reduce the occurrence of complicated peaks and dips in the filter using the head related transfer function, and more to changes in the position of the listener (dummy head) 100. It is presumed to be more robust, and it is considered that the localization feeling that the sound is output from the rear is improved over (1).

  According to the present invention, the rear virtual sound image localization is more reliably realized by outputting the sound of the rear sound input channel from the front speaker. In addition, it does not become sensitive to the orientation of the listener's face, and even if the listener moves back and forth with respect to the speaker, the sense that sound is output from the back is not impaired.

<Outline of this embodiment>
The outline of the sound image localization apparatus of the present embodiment will be described with reference to FIGS. FIG. 1 is an internal configuration diagram of the apparatus according to the present embodiment. As shown in the right of FIG. 1, it is assumed that the Lch speaker FL and the Rch speaker FR are actually provided diagonally forward in the direction 103 of the listener's face of the listener (dummy head) 100. As shown on the left side of the DSP 10, the output system outputs the front audio input channels Lch and Rch decoded by the decoder 14. Also, the left and right audio input channels LSch and RSch are input to the post-processing DSP 13. These audio input channels, which are signal-processed by the rear localization adding unit 131, are added to the front audio input channels Lch and Rch by the adding units 135A and 135B. Thus, sound image localization of the rear virtual speakers VL and VR is realized (hereinafter referred to as “addition of rear localization”). Note that the sound images are localized using the rear speakers VL and VR as virtual speakers because the output of multi-channel sound from a real speaker is large and not always practical.

  In order to realize such rear virtual sound image localization, the apparatus of this embodiment uses a model head-related transfer function that simulates the propagation characteristics from these speakers to both ears. The apparatus of this embodiment is characterized by the rear localization adding unit 131. The conventional apparatus is provided with a crosstalk cancellation correction circuit (see Patent Document 1) that cancels propagation characteristics from the speakers FL and FR to both ears M1 and M2. However, the apparatus according to the present embodiment adds rear localization. The unit 131 processes the equivalent of this crosstalk cancellation correction at the same time.

  As shown in FIG. 2 (a diagram representing a virtual sound source setting method and the like), in the apparatus of the present embodiment, the virtual speakers VL and VR are symmetrical with respect to the speakers FL and FR actually installed in front and the center line 104. Set to a correct position.

  As shown in FIG. 3, the gains of head-related transfer functions LD (ω) and LC (ω) simulating spatial propagation from the front speakers FL and FR to both ears are used as filters for the rear localization adding unit 131 ( ω is an angular frequency. The same applies hereinafter.), the gain of head transfer functions RealLD (ω) and RealRD (ω) simulating spatial propagation from the rear virtual speakers VL and VR to both ears is the angular frequency. A filter having a characteristic divided for each ω, that is, for each frequency, converted into an impulse response is used. The rear localization adding unit 131 outputs the filter through the rear audio input channel. By convolving a filter that has been calculated to divide the gain in this way, it is presumed that an effect similar to the crosstalk cancellation that cancels the propagation characteristics from the front speakers FL and FR to both ears M1 and M2 occurs.

<Configuration of sound image localization apparatus of this embodiment>
A sound image localization apparatus of the apparatus according to the present embodiment will be described with reference to FIG. FIG. 1 is an internal configuration diagram of the apparatus according to the present embodiment as described above. The sound image localization apparatus of this embodiment includes a DSP 10 that inputs various sources and processes the input. The sound image localization apparatus according to the present embodiment includes a controller 32, a user interface 33, and a memory 31. Furthermore, the sound image localization apparatus of the present embodiment includes a D / A converter 22 that converts the digital audio signal output of the DSP 10 into an analog signal, an electronic volume 41 that adjusts the volume of the audio output of the D / A converter 22, and an electronic And a power amplifier 42 for amplifying an audio signal passing through the volume 41. In addition, the speakers FL and FR are provided outside the sound image localization apparatus of the present embodiment, and the output of the power amplifier 42 is converted into sound and output to the listener (dummy head) 100. Each configuration will be described below.

  A DSP 10 shown in FIG. 1 is configured by a DSP (Digital Signal Processor), and includes a decoder 14 that decodes an input signal and a post-processing DSP 13 that processes the output signal. The decoder 14 inputs and decodes various inputs such as a bit stream of a digital audio signal, a multi PCM, and a multi bit stream. The decoder 14 outputs surround audio inputs, that is, front left and right audio input channels Lch and Rch, front center channel Cch, and rear left and right audio input channels LSch and RSch.

  The post-processing DSP 13 in FIG. 1 includes at least a rear localization adding unit 131 and addition units 135A and 135B in the rear audio input channel, and processes and outputs the surround audio input input by the decoder 14. Here, in the apparatus of this embodiment, as shown in FIG. 1, only the front speakers FL and FR are actually installed. The DSP 10 uses the rear speakers VL and VR as virtual speakers, and synthesizes the audio input of the rear channel with the audio outputs of the speakers FL and FR by the adders 135A and B to localize the sound image. Also, the center channel audio signal input Cch is distributed to Lch and Rch and synthesized by the adding sections 135A and B. The reason for mixing down as described above is that, as described above, it is large-scale and not always practical to output multi-channel audio with an actual speaker.

  The rear localization adding unit 131 includes filters 131LD, LC, RC, and RD and addition units 131L and 131R in order to localize the sound image using the rear audio input channels LSch and RSch as virtual speakers. The functional units of the filters 131LD, LC, RC, and RD are configured by a ROM 31 inside or outside the DSP and a calculation unit for convolution calculation in terms of mounting. The parameters of the FIR filter are stored in the ROM, and the calculation unit for the convolution operation performs the convolution operation of the FIR filter read from the ROM on the left and right audio input channels Lsch and Rsch for the rear in the DSP. Adders 131L and 131R add the outputs of filters 131LD, LC, RC, and RD.

The filters 131LD, LC, RC, and RD of the rear localization adding unit 131 perform head localization simulating spatial propagation from the rear virtual speaker to both ears in order to localize the sound image using the rear audio input channels LSch and RSch as virtual speakers. The function gain is divided by the head-related transfer function that simulates the spatial propagation from the front speakers FL, FR to both ears (this division is divided for each angular frequency ω, that is, for each frequency). (The details will be described later with reference to FIG. 3). Further, as shown in FIG. 1, the outputs of the filters 131LC and RC are multiplied by −1 and output. In other words, the output of the filters 131LD and RD is out of phase.
The functional blocks of the addition units 131L and 131R in FIG. 1 include a calculation unit that synthesizes the outputs of the filters 131LD, LC, RC, and RD, and outputs the outputs to the addition units 135A and 135B. The calculation for multiplying the output of the filters 131LC and RC by −1 may be performed by subtracting by the adders 131L and 131R.

  In order to synthesize the output signal of the rear localization adding unit 131 and the center channel audio signal input Cch, the adding units 135A and 135B of FIG. 1 include a calculating unit that adds signal outputs, and this calculating unit outputs the output to D / A converter 22 to output.

  The controller 32 in FIG. 1 controls the operation inside the post-processing DSP 13 in response to an instruction from the user interface 33. The memory 31 stores various control data for controlling the post-processing DSP 13. For example, the FIR parameter of the filter of the rear localization adding unit 131 is stored. The user interface 33 includes an operator and a GUI, and sends an instruction to the controller 32.

The D / A converter 22 of FIG. 1 includes a D / A converter IC, and converts a digital audio signal into an analog signal.
The electronic volume 41 is constituted by an electronic volume control IC or the like, adjusts the volume of the output of the D / A converter 22, and outputs it to the power amplifier 42. The power amplifier 42 amplifies the analog output of the electronic volume 41 and outputs it to the speakers FL and FR.

<Setting of virtual sound source of apparatus of this embodiment>
The setting of the virtual sound source of the apparatus according to the present embodiment will be described with reference to FIG. FIG. 2 is a diagram showing this setting method and the definition of the head-related transfer function used in the apparatus of this embodiment. As described above, in the apparatus of this embodiment, the rear audio input channel is localized as a virtual sound source. As shown in FIG. 2, in this embodiment, the virtual speakers VL and VR are set to positions that are line-symmetric with respect to the speakers FL and FR and the center line 104. Here, the center line 104 passes through the center of the listener 100 and is in the horizontal direction of the listener 100.

  As shown in FIG. 2, if the virtual speakers VL and VR are set at positions symmetrical to the speakers FL and FR with respect to the center line 104 in the horizontal direction of the listener, the following advantages are obtained. That is, since the propagation distances from the front speakers FL, FR and the rear virtual speakers VL, VR are the same, the phase difference due to the propagation time difference from both the front and rear, and the volume difference due to the propagation distance difference are substantially equal. Since the angles at which sound is incident are the same at the front and rear, the difference in the degree of interference between the heads can be reduced. As a result, it is presumed that complicated peaks and dips are reduced in the filter of the rear localization adding unit 131 and the position of the listener (dummy head) 100 is robust.

  In the apparatus of the present embodiment, the left and right front speakers FL and FR and the rear virtual speakers VL and VR are set at symmetrical positions with respect to the straight line of the listener's face direction 103, so The part transfer function can be the same. As a result, it is estimated that complicated peaks and dips are reduced in the frequency characteristics of the signal that has passed through the filter of the rear localization adding unit 131, and that the position of the listener (dummy head) 100 is robust. Is done.

<Setting of filter of rear localization adding unit of apparatus of this embodiment>
A filter setting method of the rear localization adding unit 131 will be described with reference to FIG. 2, FIG. 3, and FIG.

  As shown in FIG. 2, head-related transfer functions from the front speakers FL and FR and the rear virtual speakers VL and VR to both ears M1 and M2 are defined. As shown in FIG. 2, the head-related transfer function of the path propagating from the speaker to the ear on the near side is D (Direct), and the head-related transfer function of the path propagating from the speaker to the ear on the far side is C (Cross). I will write the letters. In addition, the head-related transfer function of the path propagating from the rear is expressed with a suffix “Rear”. Further, the head-related transfer function that is propagated from the diagonal left is represented as L (Left), and the one that is propagated from the right is represented as R (Right). For example, the head-related transfer function of the path 102LC diagonally backward to the left is set to RealRC (ω) (as described above, ω is an angular frequency, and so on). In both cases, these are model head-related transfer functions as described at the beginning. As this model head-related transfer function, actual measurement data is published, and the data can be used.

  A specific value of the filter of the rear localization adding unit 131 will be described with reference to FIG. FIG. 3 is a diagram showing this setting, and only the portion of the rear localization adding part 131 of FIG. 1 is enlarged. As shown in FIG. 3, each filter of the rear localization adding unit 131 takes a ratio of gains of head related transfer functions that are propagated from symmetrical positions with respect to the center line 104 in the left-right direction of the listener ( (See definition of head-related transfer function in FIG. 2). Here, the symbol “/” in the filter 131 in FIG. 3 represents the gain division at each angular frequency (the value divided by this is the difference in dB value in dB value (logarithmic notation)). Value). Although the filters 131LD, LC, RC, and RD in FIG. 3 are represented by frequency characteristics, in actuality, since time series data is input as a digital audio signal, the frequency characteristics of this gain difference are converted into FIR filters. Using this, the filter is convolved with the input signal.

  In FIG. 2, if the positions of the virtual sound source and the speakers FL and FR are set symmetrically with respect to the line of the listener's face 103, the head-related transfer function can be regarded as being symmetrical. The filters 131LD and 131RD and the filters 131LC and 131RC are the same.

  A specific example of the filter of the rear localization adding unit 131 will be described with reference to FIG. FIG. 4 shows a specific example of this. In the example of FIG. 4, the filters are shown when the positions of the speakers FL, FR, virtual sound sources VL, VR are set symmetrically with respect to the line of the listener's face direction 103 (see FIG. 3). The filters 131LD and 131RD and the filters 131LC and 131RC have the same frequency characteristics. In FIG. 4A, the filters 131LD and 131RD are represented by a curve 53. In FIG. 4B, the filters 131LC and 131RC are represented by a curve 56.

  In the example of FIG. 4, the setting angles of the speakers are 30 degrees from the listener's face orientation 103 for the front FL and FR, and 150 degrees for the rear virtual speakers VL and VR. With this setting, the speakers FL and FR and the virtual sound sources VL and VR are line-symmetric with respect to the center line 104 shown in FIG.

  As shown in FIG. 4A, the frequency responses of the filters 131LD and 131RD represented by the curve 53 are the head-related transfer functions LD (ω) and RD (ω) (LD (ω) = RD ( ω)) and the frequency response gain division of the head-related transfer functions RealLD (ω) and RealRD (ω) (RearLD (ω) = RealRD (ω)) represented by the curve 52 (this division is expressed as a dB value ( In the logarithmic notation), it is the difference value of the dB value. Similarly, in the cross direction filters 131LC and 131RC (shown by the curve 56) in FIG. 4B, the gains of the curve 54 and the curve 55 are divided. Note that these head-related transfer functions are those corresponding to the above-mentioned speaker installation angles.

  Implementation of the filter shown in FIG. 4 will be described. The gain division value as shown in FIG. 4 is calculated, and the filter of the rear localization adding unit 131 is determined in advance as a factory setting value. Then, it is stored in the memory 31 of FIG. 1 as FIR filter parameters. For this parameter, a filter may be set corresponding to the rotation angle from the orientation 103 of the listener's face in a plurality of patterns. For example, the angle can be selected from the user interface 33 according to the installation angle of the speaker installed by the user. Then, the controller 32 reads out the filter coefficient corresponding to this angle as the control parameter of the rear localization adding unit 131 and instructs the rear localization adding unit 131. As described with reference to FIG. 1, each filter of the rear localization adding unit 131 performs a convolution operation on the audio signals of the rear audio input channels LSch and RSch based on the FIR parameters.

  According to the apparatus of the present embodiment, it is perceived that the sound is surely output from the rear rather than the signal processing by the crosstalk cancellation based on the inverse matrix calculation even though it is output from the front speaker. This has been confirmed by the inventors' experiments. Note that it is presumed that the above-described division calculation has an effect similar to the crosstalk cancellation that cancels the propagation characteristics from the front speakers FL and FR to both ears M1 and M2.

As another expression of the invention of claim 1, the following configuration is conceivable.
(A) The present invention
The audio signal sequences of the left and right audio input channels are LSch and RSch, respectively.
The following equations using matrices as transfer functions LD (z), LC (z), RC (z), RD (z):
OutputL = LD (z) × LSch-RC (z) × RSch
OutputR = −LC (z) × LSch + RD (z) × RSch
("X" is a convolution operation, and "+" is an addition operation.)
A filter operation unit that performs a convolution operation and an addition operation using the operation of
The addition of adding the outputs of OutputL and OutputR, which are the calculation results of the filter calculation unit, to the Lch and Rch, respectively, as the audio signals of the front left and right audio input channels as they are or as the audio signals obtained by signal processing. And comprising
Let angular frequency be ω,
LD (ω), LC (ω), head right transfer functions simulating the spatial propagation from the left front speaker to the left and right ears, assuming the actual installation, and the right front speaker, assuming the actual installation RC (ω) and RD (ω) are head-related transfer functions that simulate spatial propagation from the ear to the left and right ears.
Assuming that a listener is a virtual speaker at the rear left and right symmetrical positions with respect to the front left and right speakers, a head-related transfer function simulating spatial propagation from the rear left virtual speaker to the left and right ears is expressed as VLD ( ω), VLC (ω), model head related transfer functions simulating spatial propagation from the right rear virtual speaker to the left and right ears as VRC (ω), VRD (ω),
The filter calculation unit uses the transfer function LD (z), LC (z), RC (z), and RD (z) as RLD (ω) divided by LD (ω) and RLC (ω), respectively. The gain ratio divided by LC (ω), the gain ratio obtained by dividing RRC (ω) by RC (ω), and the frequency response of the gain ratio obtained by dividing RRD (ω) by RD (ω) are converted into impulse responses. Sound image localization device used.

Internal configuration diagram of the sound image localization apparatus of the present embodiment The figure which showed the setting method of the virtual sound source of the sound image localization apparatus of this embodiment, and the definition of the head related transfer function used with the apparatus of this embodiment The figure showing the setting method of the filter of the rear localization addition part of the sound image localization apparatus of this embodiment The figure which shows the Example of the filter of the rear localization addition part of the sound image localization apparatus of this embodiment.

Explanation of symbols

1-sound image localization device 10-DSP
13-post-processing DSP
131-rear localization addition unit 131LD, LD, RC, RD-filter 131L, R-adder, 135A to D-adder 14-decoder 22-D / A converter 31-memory 32-controller 33-user interface 41- Electronic volume 42-Power amplifier 100-Listener (dummy head)
103-Direction of listener's face 104-Center line FL, FR-Speaker VL, VR-Virtual speaker M1-left ear M2-right ear

Claims (2)

LD, a head-related transfer function that simulates spatial propagation from the real speaker FL provided on the left front to the left ear.
LC, a head-related transfer function that simulates spatial propagation from the real speaker FL to the right ear,
RC, a head-related transfer function simulating spatial propagation from the real speaker FR provided on the front right to the left ear
A head-related transfer function simulating spatial propagation from the real speaker FR to the right ear is RD,
The center line of the listener's left and right direction passing through the center of the listener's head is L,
A head-related transfer function simulating spatial propagation from the virtual speaker VL to the left ear, which is set symmetrically with the real speaker FL with respect to the center line L, is RLD,
RLC, a head related transfer function simulating spatial propagation from the virtual speaker VL to the right ear,
A head-related transfer function simulating spatial propagation from the virtual speaker VR to the left ear that is set in line symmetry with the real speaker FR with respect to the straight line L is RRC,
RRD is a head related transfer function that simulates spatial propagation from the virtual speaker VR to the right ear.
An L direct output unit that outputs the audio signal of the audio input channel on the left rear by applying a filter having characteristics obtained by dividing RLD by LD;
An L cross output unit that outputs the audio signal of the audio input channel on the left rear by applying a filter having characteristics obtained by dividing RLC by LC;
An R-cross output unit that outputs the audio signal of the rear right audio input channel by applying a filter having characteristics obtained by dividing RRC by RC;
An R direct output unit that outputs the audio signal of the rear right audio input channel by applying a filter having characteristics obtained by dividing RRD by RD;
A first adder that adds the difference signal between the output signal of the L direct output unit and the output signal of the R cross output unit to the audio signal of the front left audio input channel, and outputs the difference signal;
A sound image comprising: a second addition unit that adds a difference signal between the output signal of the R direct output unit and the output signal of the L cross output unit to the audio signal of the front right audio input channel and outputs the resultant signal. Stereotaxic device.   The filter calculation unit sets both the speaker and the virtual speaker symmetrically with respect to the front direction of the listener, and sets the model head-related transfer function to LD = RD, LC = RC, RLD = RRD, RLC = The sound image localization apparatus according to claim 1, which is an RRC.

Images