JPH0816193A - Audio signal decoder - Google Patents

Abstract

(57) [Abstract] [Purpose] To make it possible to easily perform the process of changing the tempo without changing the pitch of the voice. A reader 2 for reading a subband-divided audio signal stored in a memory 1 at a timing corresponding to an input control signal s1 and an audio signal stored in the memory 1 for encoding Control signal s for reading at a time interval different from the set time interval
A control signal generator 9 for generating 1 is provided, and the audio signal is output from the memory 1 from the memory 1 at an interval different from the time interval at the time of subband division according to the control signal s1 generated by the control signal generator 9. By making it possible to extract, it is possible to synthesize the audio signal with a tempo different from the set tempo when sub-band splitting while maintaining the waveform of the original audio signal. It is possible to change only the tempo without changing the pitch of the reproduced voice without performing complicated processing for analyzing the waveform of the signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal decoder, and more particularly, it is suitable for use in an audio signal decoder adapted to decode a subband-divided audio signal.

[0002]

2. Description of the Related Art In recent years, a technique called MPEG has been developed as one of such compression techniques for digital audio signals. This MPEG audio signal compression technique is standardized by ISO / IEC (International Organization for Standardization / International Electrotechnical Commission), and includes both MPEG1 and MPEG2. Below, these are combined and M
It is called the PEG standard.

The principle of audio signal compression adopted in the MPEG standard will be described below. For example, a digital audio signal for one frame consisting of 384 samples is divided into 32 subbands by a subband filter,
Obtain 12 sample values for each subband. Then, the scale factor (SF) is calculated from the maximum amplitude of each subband.
And each data is normalized by the scale factor (SF) of its subband. Also, the bit amount per frame is calculated from the transmission rate and the sampling rate.

For the sub-band-divided original sampling data, the minimum and minimum hearing characteristics and masking characteristics by the human psychoacoustic model (the former characteristics that human hearing is not very sensitive in low and high frequency regions, In the latter case, the quantization level (Allocation) of each sub-band is set so as to match the bit amount of the above-mentioned one frame, on the basis of the characteristic that the hearing sensitivity decreases at a frequency near the peak of a certain frequency spectrum.

Next, according to the quantization level (Allocation) set in this way, the normalized samples for each sub-band are quantized and encoded to obtain quantized data (Sample). Then, this quantized data (Sample) is made into one data stream together with the scale factor (SF) and the quantization level (Allocation) described above, and a compressed audio signal is generated.

Here, the above-mentioned quantization level (Allocatio
n) is data used when performing an inverse quantization operation at the time of decoding, and represents how many bits are allocated to each sample included in one subband.

The scale factor (SF) is data used when performing an inverse scaling operation at the time of decoding and gives an approximate output level. For example, it corresponds to the exponent part of a floating point number used in an electronic computer.

Further, the quantized data (Sample) is data used for the inverse quantization operation at the time of decoding and gives a fine value of the output data. For example, it corresponds to the mantissa part of a floating point number used in an electronic computer.

Decoding of the data compressed and coded as described above is performed by the reverse operation of this compression / coding. The principle of this decoding will be described with reference to the configuration diagram of the conventional audio signal decoder shown in FIG. The sampling frequency adopted is 44.1 KHz.

In FIG. 8, first, the data analyzer 81 causes the data stream D _S of the target compressed audio signal.
Quantization level (Allocation), scale factor (SF) and quantized data (Sample) are decomposed and extracted from.

Quantization level (Allocation) and quantized data (Sampl) extracted by the data analyzer 81.
e) is given to the inverse quantizer 82. The inverse quantizer 82 then extracts the quantized data (Samp
le) is the similarly extracted quantization level (Allocation)
Dequantized by.

The dequantized data Sample-Value obtained by this dequantization operation is the descaling device 83 of the next stage.
Is applied to one of the input terminals. Then, the dequantized data Sample-Value is descaled by the scale factor (SF) input from the data analyzer 81 to the other input terminal, and the subband information S _j (j is 0 to 31).
Of 32 sub-band numbers) is obtained.

Next, the synthesizer synthesizer 84 subjects the subband information S _j obtained by the above inverse quantization operation and inverse scaling operation to synthesizer synthesis processing. As a result, the waveform information V _ik of the synthesizer synthesized wave shown in (Equation 1) is obtained from the frequency of each subband and the phase of the target time.

[0014]

[Equation 1]

In (Equation 1), i represents a parameter (sample number) in the time axis direction and takes a value of 0-63. Further, k represents the number of synthesizer synthesis data to be synthesized by the polyphase filter 85 at the next stage, and is 0 to 1
Takes a value of 5.

That is, in the synthesizer synthesizer 84,
First, the cosine function cos ((16 + i) × (2j + 1) π / 6 is applied to the subband information S _j given from the inverse scaling unit 83.
The value of 4) is multiplied. Next, all the multiplication values for each sub-band obtained as described above are added to obtain the waveform information V
_ik is required.

The value of the waveform information V _ik corresponds to 1/2 wavelength of the synthesizer composite wave, which is one cycle of 128 samples, that is, the value of 64 samples, and this value is obtained by the polyphase filter 85 in the subsequent stage. Since the positive and negative inversions are combined, the value corresponds to one wavelength, that is, 128 sample values. Therefore, the period f of the data synthesized here is f = 44.1K / 128≈345 Hz.

The waveform information V _ik obtained by the synthesizer synthesizer 84 is given to the polyphase filter 85. Then, the filter processing as shown in (Equation 2) is performed using a predetermined filter function PF (k, l) to suppress the discontinuity of the signal and to shift the time back and forth.
The six synthesizer synthesized waves are added and output as a PCM audio signal X _l (l = 0 to 31) in the processing circuit (not shown) in the next stage.

[0019]

[Equation 2]

In this case, the time difference Δt of each synthesizer composite wave whose time is shifted back and forth is Δt = 32 / 44.1K when the number of samples contained in each subband divided into 32. ≈726 μsec.

By the way, when reproducing an audio signal, it may be desired to change the reproduction speed by changing the pitch of the reproduced audio. For example, in the field of karaoke,
It is desirable to be able to freely change the playing speed according to the singer's taste.

When the voice to be reproduced is a person's conversation, the reproduction speed is slowed to make the conversation easier to hear, and the reproduction speed is increased to shorten the entire reproduction time. It would be convenient if possible. Therefore, conventionally, there has been proposed a voice reproducing apparatus capable of changing the reproduction speed of a target voice.

[0023]

However, in the conventional audio reproducing apparatus, there is a problem that the pitch of the reproduced sound changes correspondingly when the tempo of the reproduced sound is changed. That is, when the tempo of the reproduced voice is slowed down, the pitch becomes low, and conversely, when the tempo is increased, the pitch becomes high.

In order to solve such a problem, conventionally, when the reproduction speed is changed, the entire reproduction time is adjusted by copying or deleting a part of the waveform of the original audio signal. Was being done.

However, in such a method, in order to copy or delete a part of the waveform, it was necessary to analyze the waveform of the original audio signal and examine its wavelength and the like. Therefore, there is a problem that the processing becomes very complicated and the processing cost becomes high. Further, by copying a part of the waveform, the reproduced voice signal becomes different from the waveform of the original voice signal, so that the sound quality is deteriorated.

The present invention has been made to solve such a problem, and makes it possible to easily perform a process of changing the tempo without changing the pitch of the voice. It is an object.

[0027]

A voice signal decoder according to the present invention is a voice signal decoder for decoding a voice signal which is subband-divided into a plurality of set frequencies. And a control signal is generated for the signal capturing means to capture the audio signal at a time interval different from the time interval set for encoding the audio signal. Control signal generating means for supplying to the signal capturing means, and variable timing synthesizing means for synthesizing and outputting the subband-divided audio signal captured by the signal capturing means from a plurality of periodic functions. The time interval of the subband-divided audio signal can be changed and decoded.

Another feature of the present invention is that, in an audio signal decoder for decoding an audio signal sub-band-divided into a plurality of set frequencies, the input audio signal is supplied with a first control signal. , And the subband-divided audio signal fetched by the signal fetching means is synthesized and output according to a periodic function according to the second control signal input. The variable timing synthesizing means and the signal capturing means generate the first control signal so as to capture the audio signal at a time interval different from the time interval set at the time of encoding the audio signal, and the signal is generated. The audio signal is supplied to the capturing means and is decoded according to a frequency different from the sampling frequency when the audio signal is subband-divided. A second control signal to generate,
A control signal generating means for supplying to the variable timing synthesizing means, so that the time interval of the sub-band divided audio signal can be changed and decoded, and the setting when the sub-band is divided. The audio signal can be decoded according to a periodic function of a frequency different from the frequency.

Still another feature of the present invention is that a window processing means for performing window processing on a voice signal obtained by the variable timing synthesis means using a predetermined window function is provided, and the window processing means is provided. The window processing performed is changed according to a predetermined control signal generated by the control signal generating means.

[0030]

According to the present invention configured as described above, it becomes possible to arbitrarily change the timing of taking in the audio signal subband-divided into a plurality of frequencies according to the control signal generated by the control signal generating means, Since the subband-divided audio signal is synthesized at a tempo different from the set tempo at the time of encoding without changing its frequency component, conventionally, the pitch of the reproduced audio is not changed. It is not necessary to perform the waveform analysis processing that was necessary to change only the tempo.

According to another feature of the present invention, the frequency at which the variable-timing synthesizing means synthesizes the audio signals subband-divided into a plurality of set frequencies is generated by the control signal generating means. Since it can be changed arbitrarily according to the control signal of, it becomes possible to synthesize the audio signal according to the changed periodic function of the frequency, and the audio signal at a frequency different from the set frequency when subband division is performed. Can be synthesized.

According to still another feature of the present invention, the window function is changed according to the predetermined control signal generated by the control signal generating means, and the optimum window function is always obtained. By performing the predetermined window processing according to the window function, noise due to the discontinuity point of the audio signal obtained by the variable timing synthesizing means, the change of the synthesis frequency in the variable timing synthesizing means, or the change of the audio signal capturing period The sound quality deterioration is suppressed.

[0033]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows an audio signal decoder, which is constructed by applying an audio signal compression and decoding technique based on the MPEG standard.

The MPEG standard is characterized in that an audio signal is divided into a plurality of subbands for compression and decoding. For example, when encoding a voice signal having a waveform as shown in FIG. 2A, first, t = t ₁ , t = t ₂ ,
The audio signal at each time t = t ₃ is cut out, and the result is shown in FIG.
A plurality of audio signals having a waveform like this are obtained.

This is because it is not practical to calculate all the regions from -∞ to + ∞ in the sub-band division processing at the time of encoding, so that a certain section of the signal to be analyzed is set. This is because it is cut out and the subsequent processing is performed assuming that the signal continues infinitely.

However, as shown in FIG. 2B, the respective audio signals cut out at each time are not always continuous at the end points of the section. Therefore, the above-mentioned discontinuity is suppressed by applying a window function having a shape illustrated in FIG. 2C to the audio signal for each time to converge the end points to a predetermined level, By obtaining the voice signal at each time having the waveform as shown in FIG. 2D, it is possible to reduce the occurrence of noise in the analysis result.

Next, the frequency component information at each time as shown in FIG. 2E is obtained by frequency-analyzing the audio signal at each time after the above window processing and dividing into subbands. As described above, t = t ₁ , t ₂ ,
By obtaining the frequency component information at each time t ₃ , ..., The subband-divided audio signal information at each time as shown in FIG. 2 (f) is obtained. This audio signal information is two-dimensional matrix information having two axes, a frequency axis and a time axis, which are encoded under the set frequency resolution and time interval.

Next, the operation for decoding the voice signal for each sub-band encoded as described above will be explained using the voice signal decoder shown in FIG. At first,
A case of reproducing an audio signal without changing the tempo, that is, a decoding operation by a normal audio decoding method will be described below with reference to the waveform diagram of FIG.

In FIG. 1, first, a data analyzer 81, an inverse quantizer 82 and an inverse scaling unit 83 having the same functions as those shown in FIG. Inverse scaling processing is performed to obtain subband-divided audio signal information as shown in FIG. The audio signal information is shown in FIG.
It is the same two-dimensional matrix information as shown in (f).

When the audio signal information for each subband (frequency component information at each time) is obtained in this manner, the write request signal s3 is given to the memory controller 10. Then, in response to the write request signal s3, a predetermined write control signal s5 is given from the memory controller 10 to the memory 1, and the above M
The audio signal information for each sub-band obtained by analyzing the PEG data is given to the memory 1 and stored therein.

The frequency component information (subband information S _j ) stored in the memory 1 is read by the variable timing frequency component reader 2 and given to the fixed-cycle synthesizer synthesizer 3 (FIG. 3 (b)). )reference).

When the frequency component information is read, the read request signal s4 is sent to the memory controller 10 in accordance with the subband information read signal s1 output from the timing generator 8 to the variable timing frequency component reader 2 at an arbitrary set cycle. Given to.

Then, in response to the read request signal s4,
A predetermined read control signal s from the memory controller 10 to the memory 1
When 6 is given, the frequency component information is read from the memory 1. In a normal voice decoding system, the cycle of the subband information read signal s1 is the same as the time interval set at the time of encoding (for example, 44.1 KH).
z cycle).

Next, the sub-band information S _j read by the variable timing frequency component reader 2 is
The signal component is given to the signal analyzer 11 and its signal component is analyzed. The analysis information obtained by this is given to the synthesizer controller 12 and the window controller 13, and the operations of the synthesizer synthesizer 3 of fixed cycle and the window processor 5 of the variable window function are controlled based on this analysis information. As a result, synthesis control according to frequency component information as described below is performed.

That is, in the fixed-cycle synthesizer synthesizer 3, the variable-timing frequency component reader 2 is used.
The sub-band information S _j read by is synthesized by the synthesizer to obtain the waveform information V _ik . This waveform information V
_ik is obtained by the arithmetic expression shown in the following (Expression 3).

[0046]

(Equation 3)

Under the normal speech decoding method, the values of M and N in the above (formula 3) are M = 16 and N = 64 as standard values, respectively, and the cosine function cos ((M + i) ×
The number of samples included in one cycle of (2j + 1) π / N) is
As in the case of (Expression 1) described above, 2N = 128.

That is, weighting is executed by the frequency component information read by the frequency component reader 2 at the variable timing by the periodic function having the same frequency interval as the frequency interval (frequency resolution) set at the time of encoding the audio signal. Then, the waveform information V _ik is synthesized. In this case, the synthesizer composite wave indicated by the waveform information _Vik is as shown in FIG.
become that way.

The waveform information V _ik obtained by the fixed-cycle synthesizer synthesizer 3 in this manner is given to the window processor 5 of the variable window function which constitutes the polyphase filter 4. Then, the window function WIN having the shape illustrated in FIG.
The window processing is performed by using (k, l) by an arithmetic expression as shown in (Expression 4). As a result, the discontinuity of the signal shown in FIG. 3C is suppressed, and the window processing information W _kl having the waveform as shown in FIG. 3E is obtained. W _kl = V _ik × WIN (k, l) (Equation 4)

Under the normal speech decoding method, the window function WIN (k, l) used in (Equation 4) is equal to the filter function PF (k, l) used at the time of encoding. Figure 2
It has the same shape as that shown in FIG.

Next, the variable timing synthesizer 6 synthesizes 16 pieces of window processing information W _kl whose time is shifted back and forth based on the following (Equation 5) to obtain the PCM voice signal X _l. To be Then, the PCM audio signal X ₁ is output to the processing circuit (not shown) in the next stage in accordance with the sampling cycle based on the D / A output timing signal s2 output from the clock generator 7.

[0052]

[Equation 4]

As a result, as shown in FIG. 3 (f), the window processing information W _{kl, which} has been obtained up to the window processing unit 5 of the variable window function and is discrete for each time, is a continuous audio signal on the time axis. It will be output as X _l . As described above, in the normal audio decoding method, the original audio signal is reproduced by performing the operation completely opposite to that at the time of encoding.

The operations of the memory controller 10, the fixed-cycle synthesizer synthesizer 3, the variable window function window processor 5, and the variable timing synthesizer 6 are performed by the clock which is a component of the control signal generator 9. It is controlled according to the D / A output timing signal s2 output from the generator 7.

Further, the D / A output timing signal s2
Is also given to the timing generator 8, which is another component of the control signal generator 9. The timing generator 8 divides the frequency of the D / A output timing signal s2 into 1 / N ₁ to generate the subband information read signal s1.

Next, the decoding operation when the tempo control is performed to reproduce the audio signal will be described below. In the present embodiment, the tempo of the reproduced voice can be freely changed by changing the time intervals of the 16 synthesizer composite waves that are synthesized in the polyphase filter 4 and are temporally before and after each other.

To change the time interval of the synthesized wave of the synthesizer, the period of the subband information read signal s1 generated by the timing generator 8 is changed to change the number of samples included in each subband, and the clock generator 7 is used. There are two methods, that is, a method of changing the cycle of the D / A output timing signal s2 generated in 1. In the latter case, the value of N in (Equation 3) is changed so that the frequency of the reproduced audio signal is fixed.

The tempo control coefficient T _x , which indicates at what speed the voice after the tempo control is reproduced at the original voice, is given by the following (formula 6). However, N ₁ and [s2] represent the number of samples included in each subband and the standard value of the period of the D / A output timing signal s2, respectively. That is, N ₁ = 32, [s2] = 44.
It is 1 KHz. Further, N ₁ 'and [s2'] represent arbitrary values obtained by changing the standard values of N ₁ and [s2] described above in order to perform tempo control.

[0059]

(Equation 5)

For example, when the number of samples is N ₁ = 32 in the normal speech decoding system and is changed to N ₁ ′ = 16, the time interval Δt ′ between the synthesized synthesizer waves shifted back and forth is Δt ′ = 16. /44.1K≈363 μsec, which is the normal time difference Δt≈726 μsec to Δt ′.
It can be changed to ≈363 μsec.

Further, when the period of the D / A output timing signal s2 is 44.1 KHz in the normal speech decoding system, but it is changed to 32 KHz, the time interval Δt 'of each synthesized synthesizer wave shifted back and forth becomes Δt'. = 32 / 32K = 1000 μsec, which is the normal time difference Δt≈726 μsec to Δt ′.
= 1000 μsec.

The decoding operation with the above tempo control will be described with reference to FIGS. 1 and 4.
It looks like this: That is, in FIG. 1, up to the point where the frequency component information for each subband is stored in the memory 1, it is the same as in the case of the normal speech decoding method. Therefore, the audio signal information shown in FIG.
Is the same as that shown in.

The frequency component information stored in the memory 1 is read by the frequency component reader 2 with variable timing.
The sub-band information read signal s1 output from the timing generator 8 is read. As described above, the period of the subband information read signal s1 is the value of the number of samples N ₁ included in each subband or D /
By changing the cycle of the A output timing signal s2, it is set to an arbitrary cycle.

As a result, the frequency component information stored in the memory 1 is read out at an interval different from the time interval set at the time of encoding. As a result, the obtained frequency component information has the original tempo t = t as shown in FIG.
Tempo different from ₁ , t ₂ , t ₃ , ... T = t ₁ ', t ₂ ',
Information having t ₃ ', ...

Next, in the fixed-cycle synthesizer synthesizer 3, synthesizer synthesis is performed according to a periodic function having the same frequency as the frequency interval set at the time of encoding, based on the arithmetic expression shown in (Equation 3). A synthesizer composite wave having a waveform like 4 (c) is obtained. The waveform information of the synthesizer composite wave obtained by changing the reading cycle of the frequency component information as described above is represented by _Vik '.

In the above (formula 3), when the tempo control is performed by changing the value of the number of samples N ₁ included in each subband, the value of N representing the cycle of the above periodic function is fixed and N = 64. On the other hand, when performing the tempo control by changing the period of the D / A output timing signal s2, the pitch is also changed at the same time, so that the tempo control alone can be performed by appropriately changing the value of N.

As is apparent from FIG. 4 (c), the synthesizer synthesized wave indicated by the waveform information V _ik 'is a waveform in which only the tempo is changed without changing the pitch of the voice during compression. There is.

The waveform information V _ik ′ obtained by the fixed-cycle synthesizer synthesizer 3 is then given to the window processor 5 of the variable window function, and the window function W having the shape illustrated in FIG.
Window processing is performed according to IN (k, l) '. The window function WIN (k, l) ′ used in this window processing can be changed in accordance with the arbitrarily changed frequency component information reading cycle, and is represented by the above equation (3). The shape is changed by changing the value of M.

In this way, when the window processing unit 5 of the variable window function performs the predetermined window processing, the window processing information W _kl 'having a waveform as shown in FIG. 4E is obtained. As a result, the noise due to the discontinuity of the window processing information W _kl 'generated by the fixed-cycle synthesizer synthesizer 3 and the deterioration of the sound quality due to the tempo change can be suppressed. Thereafter, the variable timing synthesizer 6 synthesizes the 16 pieces of window processing information W _kl ′ whose time is shifted back and forth, and outputs the PCM audio signal X _l ′ to the processing circuit (not shown) in the next stage. It

As described above, according to this embodiment, the timing of reading the frequency component information using the control signal (subband information read signal s1 and D / A output timing signal s2) generated by the control signal generator 9 is obtained. Just change
You can change only the tempo without changing the pitch of the playback sound. Therefore, it is possible to freely change the tempo of the reproduced voice with the same pitch without performing complicated processing, and it is possible to prevent the sound quality from deteriorating.

Next, another embodiment will be described with reference to FIGS. FIG. 5 is a block diagram showing the configuration of an audio signal decoder according to this embodiment, FIG. 6 is a diagram showing a specific configuration of a variable cycle synthesizer synthesizer which is a component of the above audio signal decoder, and FIG. 7 is this embodiment. 6 is a waveform diagram for explaining the operation of the audio decoding system performed by the audio signal decoder of FIG.

As is clear from FIG. 5, in this embodiment,
The fixed cycle synthesizer synthesizer 3 used in the audio signal decoder of FIG. 1 is replaced with a variable cycle synthesizer synthesizer 23 capable of arbitrarily changing the frequency at the time of synthesizer synthesis.

This variable period synthesizer synthesizer 23
Is a plurality of element signal generators 3 ₀ , 3 ₁ , 3 ₂ , ... As shown in FIG.
And one adder 31. That is, the element signal generators 3 ₀ , 3 ₁ , 3 ₂ , provided corresponding to the subband information S _j of the subband numbers 0 to 31 given from the variable-timing frequency component reader 2 are provided. Cosine function cos ((M + i) × (2j + 1) π
The value of / N) is multiplied. Next, the adder 31 adds all the element signals for each sub-band thus obtained to obtain the waveform information V _ik .

As described above, the pitch of the reproduced voice can be freely changed by changing the frequency of synthesizer synthesis in the variable-cycle synthesizer synthesizer 23 to perform pitch control.

In order to change the frequency to be synthesized by the synthesizer, the fundamental period (cosine function cos ((M + i) ×) of the element signal generators 3 ₀ , 3 ₁ , 3 ₂ , ... Of the variable period synthesizer synthesizer 23 is changed.
(Represented by 2N using the value of N in (2j + 1) π / N)) and a method of changing the cycle of the D / A output timing signal s2 generated by the clock generator 7. There are two methods.

A pitch control coefficient P _x representing how many times the pitch of the voice after the pitch control is reproduced as compared with the original voice is given by the following (formula 7). However, (Equation 7) N and [s2] in, each element signal generator 3 _0, 3
_1, 3 _2, ... standard value of the period values and D / A output timing signal s2 representative of a half cycle of the respectively. That is, N = 64 and [s2] = 44.1 KHz. Further, N ′ and [s2 ′] represent arbitrary values obtained by changing the standard values of N and [s2] described above to perform pitch control.

[0077]

(Equation 6)

For example, N = 64 in the normal voice decoding system.
When N ′ = 32 is changed, the period f ′ of the synthesizer synthesized wave synthesized by the variable period synthesizer synthesizer 23 is f ′ = 44.1K / 64≈689 Hz, and the normal period f≈ From 345 Hz, f'≈689H
Can be changed to z.

Further, in the normal voice decoding system, the cycle (sampling frequency) of the D / A output timing signal s2 is 4
If the frequency of 4.1 KHz is changed to 32 KHz, the period f ′ of the synthesizer synthesized wave synthesized by the variable period synthesizer synthesizer 23 is f ′ = 32 K / 128≈250 Hz, which is the normal period f≈345 Hz to f '≒ 250H
Can be changed to z.

The decoding operation with tempo control and pitch control will be described below with reference to FIGS. 5 and 7. That is, FIG.
In the above, up to the point where the frequency component information for each subband is read by the frequency component reader 2 with variable timing, it is the same as the audio decoding method in the case of performing only tempo control. As a result, as shown in FIGS.
The audio signal information and the frequency component information as shown in are obtained.

Next, the frequency component information (subband information S _j ) read by the variable timing frequency component reader 2 is used as a variable period synthesizer synthesizer 23.
Given to the synthesizer.

That is, in the variable-cycle synthesizer synthesizer 23, by changing the value of N in the arithmetic expression shown in (Equation 3) to an arbitrary value, a frequency different from the frequency interval set at the time of encoding. The periodic function of is generated. Then, synthesizer synthesis is performed based on this periodic function, and the waveform information V _ik ″ of the synthesizer synthesized wave having the waveform as shown in FIG. 7C is obtained. This synthesizer synthesized wave has a waveform in which the tempo of the voice during compression is changed and the pitch is changed.

Then, the waveform information V _ik ″ obtained by the variable-cycle synthesizer synthesizer 23 is given to the window processor 5 of the variable window function, and the window function WIN having the shape illustrated in FIG. Windowing is performed according to (k, l) ''. The window function WIN (k, l) '' used in this window process is
The shape is changed by changing the value of M in the arithmetic expression shown by.

When the window processing unit 5 of the variable window function performs the predetermined window processing in this manner, window processing information W _kl ″ having a waveform as shown in FIG. _7E is obtained. As a result, the window processing information W _kl ″ generated by the variable-cycle synthesizer synthesizer 23 is generated.
The noise due to the influence of the discontinuity point and the deterioration of the sound quality due to the change of the tempo and the pitch can be suppressed. After that, the variable timing synthesizer 6 synthesizes 16 pieces of window processing information W _kl ″ whose time is shifted back and forth, and FIG.
The PCM audio signal X _l ″ having the waveform as shown in (f) is output to the processing circuit (not shown) in the next stage.

As described above, according to this embodiment, the control signals (subband information read signal s1 and D / A output timing signal s2) generated by the control signal generator 9 are used to synthesize each synthesizer composite wave. The tempo and pitch of the reproduced voice can be freely changed only by changing the time interval and the frequency for synthesizer synthesis. Therefore,
It is possible to freely change the tempo and pitch of the reproduced sound without performing complicated processing, and it is also possible to prevent the sound quality from deteriorating.

Further, since the tempo and the pitch can be controlled independently of each other, the combination of the tempo and the pitch of the reproduced voice can be freely selected. For example, it is possible to increase the tempo even though the sound is reproduced in the low tone, or to decrease the tempo while reproducing the sound in the high tone. This cannot be done by the conventional audio reproducing device.

As described above, the parameters of the tempo control and the pitch control are N, s.
There are three, 2, N ₁ . Then, the tempo control is performed by the subband information read signal s1 (= [s2]
This can be done by changing N ₁ ). The pitch control can be performed by changing the value of the sample number N included in the D / A output timing signal s2 or each subband.

Below, examples of the above-mentioned three parameters used in each of the normal speech decoding method, the speech decoding method when the tempo is changed, and the speech decoding method when the tempo / pitch is changed are described for reference. Tables 1 and 2 below summarize the results.

[0089]

[Table 1]

[0090]

[Table 2]

In the above embodiments, the case where the MPEG standard audio signal decoder is used has been described. However, other than that, other devices may be used as long as the device adopts the method of sub-band dividing the audio signal. The present invention can be applied to the decoder of. Further, even in a device that does not adopt the sub-band division method, the present invention can be similarly applied by adopting a mechanism of sub-band dividing an audio signal.

In the text, the pitch of the sound is described as the pitch and the reproduction speed is described as the tempo, but this is commonly used in karaoke and the like.

[0093]

As described above, according to the present invention, the signal capturing means for capturing the input audio signal at the timing corresponding to the input control signal, and the audio signal encoded by the signal capturing means. Control signal generating means for generating the control signal so as to capture the audio signal at a time interval different from the time interval set at time and supplying the signal to the signal capturing means is provided and generated by the control signal generating means. Depending on the control signal, the time interval of the subband-divided audio signal can be arbitrarily changed and synthesized, so that the subband-divided audio at a tempo different from the tempo set at the time of encoding. The signals can be synthesized and their frequency components can be kept unchanged, so that the reproduced sound can be reproduced without complicated processing to analyze the waveform of the original audio signal. It is possible to change only the tempo without changing the height. Moreover, as described above, since the audio signals can be combined while maintaining the original waveform, it is possible to reduce the sound quality deterioration of the reproduced audio signal.

According to another feature of the present invention, the control signal generating means is configured to generate a control signal having a frequency different from the set frequency when the sub-band is divided, and when the sub-band is divided, Since it is possible to synthesize an audio signal with an arbitrary frequency different from the set frequency of, it is possible to obtain an audio signal having a frequency component different from the above set frequency, and to adjust the pitch of the reproduced audio signal. You can change it freely.

According to still another feature of the present invention, window processing means for performing window processing on the audio signal obtained by the variable timing synthesizing means using a predetermined window function is provided, and the window processing means performs the window processing. Since the window processing to be performed is changed according to the predetermined control signal generated by the control signal generating means, it is possible to always perform the window processing by using the optimum window function, and to obtain it by the synthesizing means. It is possible to sufficiently suppress the noise due to the discontinuity point of the audio signal to be generated, and the deterioration of the sound quality due to the change of the synthetic frequency in the synthesizing means or the change of the audio signal capturing period.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an audio signal decoder which is an embodiment of the present invention.

FIG. 2 is a signal waveform diagram for explaining a procedure of generating a subband-divided audio signal.

FIG. 3 is a signal waveform diagram for explaining an operation of a normal speech decoding system.

FIG. 4 is a signal waveform diagram for explaining the operation of the audio decoding system when tempo control is performed.

FIG. 5 is a block diagram showing a configuration of an audio signal decoder according to another embodiment.

FIG. 6 is a diagram showing a specific configuration of a variable cycle synthesizer synthesizer.

FIG. 7 is a signal waveform diagram for explaining the operation of the speech decoding system when the tempo / pitch is changed.

FIG. 8 is a block diagram showing a configuration of a conventional audio signal decoder.

[Explanation of symbols]

1 memory 2 variable timing frequency component reader 3 fixed period synthesizer synthesizer 3 ₀ , 3 ₁ , 3 ₂ , 3 ₃ element signal generator 4 polyphase filter 5 variable window function window processor 6 variable timing synthesizer 7 Clock Generator 8 Timing Generator 9 Control Signal Generator 10 Memory Controller 11 Signal Analyzer 12 Synthesizer Controller 13 Window Controller 23 Variable-cycle Synthesizer Synthesizer

Claims (8)

[Claims] 1. An audio signal decoder for decoding an audio signal sub-band-divided into a plurality of set frequencies, and signal acquisition means for acquiring the input audio signal at a timing according to an input control signal, Control signal generation for generating the control signal so that the audio signal is captured by the signal capturing means at a time interval different from the time interval set at the time of encoding, and is supplied to the signal capturing means. Means and variable timing synthesizing means for synthesizing and outputting the subband-divided audio signal captured by the signal-capturing means from a plurality of periodic functions, the time interval of the subband-divided audio signal An audio signal decoder characterized by being able to change and decode. 2. The audio signal decoder according to claim 1, wherein a buffer memory is provided in front of the signal receiving means, and the signal receiving means receives the audio signal temporarily stored in the buffer memory according to the control signal. An audio signal decoder which is read out and supplied to the variable timing synthesizing means. 3. An audio signal decoder for decoding an audio signal subband-divided into a plurality of set frequencies, wherein a composite wave synthesized based on the subband-divided audio signal is generated according to an input control signal. Variable timing synthesizing means for synthesizing and outputting in accordance with the sampling frequency, and control signal generating means for generating a control signal of a frequency different from the sampling frequency when the audio signal is subband-divided and deriving it to the variable timing synthesizing means. An audio signal decoder comprising: and capable of changing the time interval of the sub-band-divided audio signal for decoding. 4. An audio signal decoder for decoding an audio signal sub-band-divided into a plurality of set frequencies, wherein the input audio signal is taken in at a timing corresponding to an input first control signal. Means, variable timing synthesizing means for synthesizing and outputting the subband-divided audio signal fetched by the signal fetching means in accordance with a periodic function according to an input second control signal, and the signal fetching means. The first control signal is generated and supplied to the signal capturing means so that the audio signal is captured by the capturing means at a time interval different from the time interval set at the time of encoding. Generating a second control signal that causes the audio signal to be decoded according to a frequency different from the sampling frequency when the audio signal was subband divided. And a control signal generating means for supplying the variable timing synthesizing means so that the time interval of the subband-divided audio signal can be changed and decoded. An audio signal decoder, wherein the audio signal can be decoded according to a periodic function of a frequency different from a set frequency. 5. The audio signal decoder according to claim 4, wherein a buffer memory is provided in front of the signal receiving means, and the signal receiving means receives the audio signal temporarily stored in the buffer memory as the first signal. An audio signal decoder which is read according to a control signal and is supplied to the variable timing synthesizing means. 6. The variable timing synthesizing means comprises synthesizing means for performing synthesizer synthesizing based on a predetermined synthesizing arithmetic expression, and the synthesizing arithmetic operation according to a second control signal inputted to the synthesizing means. 6. The audio signal decoder according to claim 3, 4 or 5, wherein the audio signal can be decoded according to a frequency different from a set frequency when the subband is divided by changing the expression. 7. A window processing means for performing window processing on a voice signal obtained by the variable timing synthesis means using a predetermined window function is provided, and the window processing performed by the window processing means is performed by the control signal generation means. The audio signal decoder according to claim 1, 2 or 3, wherein the audio signal decoder is changed according to the control signal generated in (1). 8. A window processing means for performing window processing on the audio signal obtained by the variable timing synthesis means using a predetermined window function is provided, and the window processing performed by the window processing means is performed by the control signal generation means. 5. The change is made according to the first control signal and the second control signal generated by
The audio signal decoder described in 5 or 6.