JP2587591B2 - Audio / musical sound band division encoding / decoding device - Google Patents

Abstract

PURPOSE:To decrease the scale and increase the speed by dividing a digital signal of a speech and a musical sound into plural bands through a band-pass filter group, and sampling the respective band digital signals down to a Nyquist rate and then performing adaptive conversion and encoding. CONSTITUTION:The speech and musical sound signal is divided by the low-pass filter 1 and high-pass filter 2 and made into a sampling signal of 8KHz through down samplers 3 and 4, and the signal is stored in buffers 5 and 6 in the form of blocks of 32msec; and ATC encoders 7 and 8 perform the adaptive conversion and encoding. The data signals of the low and high bands and auxiliary information are multiplexed by a multiplexer 11 and sent out to a transmission channel. In the band division encoding, complex arithmetic points of one block are halved in number from 2N=512 at the time of 16KHz sampling to N=256 and the complex arithmetic quantity of one block is therefore halved from Nlog2N to 2X(NX2)log2(N/2).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice / musical sound band division encoding / decoding apparatus.

[0002]

2. Description of the Related Art Conventionally, a band filter and an ADPCM (adaptive difference PCM) encoder shown in FIGS.
An apparatus for performing sub-band coding using a decoder is known. In the encoding of this apparatus (see FIG. 2A), an input digital signal representing a voice or a musical tone is divided into two bands through a low-pass filter 41 and a high-pass filter 42, and then is divided into down-samplers 43 and 43, respectively. 2 for 44
The down-sampled signal is encoded by the ADPCM encoders 45 and 46 into an ADPCM signal, and then combined again by the multiplexer 47 and sent out to the transmission channel. In the decoding of this device (see FIG. 2 (b)), the signal input from the transmission channel is divided into two bands, high and low, by way of a demultiplexer 48, and then the ADPCM decoder 49, The data is decoded at 50, and subsequently, each of the upsamplers 51 and 52 adds "0" to each point and performs upsampling of 1: 2. After passing through the high-pass filter 53 and the low-pass filter 54, respectively, it is added again. The reproduced signal is obtained by being combined by the device 55. This sub-band coding apparatus uses the ADPCM coding to optimally change the quantization width when coding the difference between adjacent sample values. This has the advantage that more information can be allocated to the required bandwidth.

[0003] However, in recent years, the above-mentioned sub-band coding (SBC) apparatus is more advantageous and has much higher flexibility.
An adaptive transform coding (ATC) device as shown in (a) and (b) has appeared.

In the encoding of this apparatus (see FIG. 3 (a)), an input digital signal representing a voice or a musical tone is stored in a buffer 60 as one block every short period (20 to 30 msec) in which the signal is substantially stationary. Then, a discrete cosine transform (DCT) is performed on the frequency domain by the DCT calculator 61, and the transform coefficient (frequency component) is quantized and coded by the quantizer / coder 62. At this time, the auxiliary information calculation / quantizer 63
The transform coefficient output from the DCT calculator 61 is divided into several bands, the average power of each band is obtained, and this is encoded as auxiliary information and output to the auxiliary information interpolator 64. The auxiliary information interpolator 64 obtains spectrum envelope information by interpolating the auxiliary information, and outputs it to the bit allocation / step width calculator 65. The bit allocation / step width calculator 65 A control signal is output to the quantizer / encoder 62 based on the envelope information, and the control signal is used to adaptively control the quantization width in the quantizer / encoder 62 and allocate the optimal number of quantization bits to each transform coefficient. Is performed. Next, the quantized and coded data signal is combined with the auxiliary information by the multiplexer 66 and transmitted to the transmission channel.
In a transmission digital signal obtained by such quantization and encoding, the quantization noise becomes uniform on the frequency axis, and the distortion to the input digital signal is minimized.

In the decoding of the ATC device (see FIG. 3 (b)), a transmission digital signal from a transmission channel is divided into a data signal and auxiliary information via a demultiplexer 67, and the auxiliary information is converted into an auxiliary information interpolator. At 68, the spectral envelope is obtained and input to the bit allocation / step width calculator 69, while the data signal is input to the decoder 70, and the bit allocation / step width calculator 69 is input.
Is decoded into a digital signal representing DCT coefficients based on the control signal from. Next, the decoded digital signal is inversely converted from a spectrum into a waveform by an inverse DCT calculator 71, and is reproduced as an original input digital signal for each block through a buffer 72.

The ATC apparatus is generally used when sampling a voice / tone signal having a maximum frequency of 4 kHz or less at a sampling frequency of 8 kHz and transmitting it at a bit rate of about 16 to 32 kbps. In that case, DC
If one block of the T conversion is, for example, 32 msec, N = 2
Requires a discrete cosine transform operation on 56 sample values,
This requires N ² = 65536 complex multiplications and additions, which increases the computation time. Therefore, a fast Fourier transform (FFT) was devised to perform this operation at high speed.
According to this method, the number of complex operations is calculated as (N / 2) log ₂ (N / 2)
…… (1), that is, it can be reduced to approximately 1/73 up to 896 times (J. Makhoul, “A Fast Cosine Transform in on
e and two dimensions, IEEE ASSP vol.28, No.1.Feb,
1980, pp27-34). Therefore, this FFT is represented by A in FIG.
The size of the DCT operation unit 61 and the inverse DCT operation unit 71, and thus the size of the ATC device, are reduced in the TC device. Since such an ATC device performs an operation for conversion, it is larger in scale than the subband coding device shown in FIG. 3, but has a high degree of freedom as described above.
Since high-quality transmission can be performed at a bit rate of 16 to 32 kbps, the bit rate tends to be frequently used in recent years.

[0007]

However, when the above-mentioned conventional ATC device is applied to the encoding of a speech / tone signal having a band of about 7 kHz, the following problems become apparent. Was. That is, such a voice / tone signal needs to be sampled at a sampling frequency of 16 kHz, and one block of DCT conversion is performed for 32 ms as described above.
Assuming that ec, DCT transform operation is required for a sample value sequence of N = 512, which is twice the conventional value. Then, DCT
Even if FFT is used for the conversion operation, the amount of complex operation reaches approximately 2.3 times the conventional value, and it takes a long time to perform the operation or D
There is a problem that the scale of the CT calculator 61 and thus the size of the ATC device becomes extremely large.

Accordingly, a first object of the present invention is to perform encoding / decoding of a speech / tone signal having a band of 7 kHz or more at a higher speed than a conventional ATC device or a smaller scale than a conventional ATC device. An object of the present invention is to provide a speech / music sound band division encoding / decoding apparatus which can be performed by the apparatus.

A second object of the present invention is to provide a voice / musical sound signal which can perform encoding / decoding of a voice / musical sound signal over a wide band with higher performance even at a device scale and bit rate similar to those of a conventional SBC device. An object of the present invention is to provide a band division coding / decoding device.

[0010]

In order to achieve the above object, a voice / musical sound band division encoding / decoding apparatus according to the present invention is characterized in that the encoding side converts at least first and second input digital signals of voice / musical sound. First and second bandpass filters for dividing into digital signals of two bands, first adaptive transform coding means for adaptively transforming the digital signals of the first band, and digital signals of the second band A second adaptive transform coding means for adaptively transforming
And the second band of digital signals.
A single transform coefficient calculating means for calculating transform coefficients for both the first and second adaptive transform coding means, and at least the first and second digital signals of the first and second bands, And a single bit allocation calculating means for performing bit allocation in consideration of the characteristics of the digital signal of the band and outputting it to the first and second adaptive transform coding means. Separating means for dividing the data signal into at least the first and second bands;
First adaptive decoding means for adaptively transforming and decoding the data signal of the second band, second adaptive decoding means for adaptively converting and decoding the data signal of the second band, and at least the first adaptive decoding means.
A single transform coefficient computing means for computing transform coefficients for both the first and second adaptive transform decoding means, shared by the data signals of the first and second bands, and at least the first and second transform coefficients.
, Which is shared by the data signals of the first and second bands, performs bit allocation in consideration of the characteristics of the data signals of the first and second bands, and outputs the bits to the first and second adaptive conversion decoding means. And a bit allocation calculating means.

In the above-mentioned audio / musical sound band division encoding / decoding apparatus, each digital signal output from the first and second bandpass filters on the encoding side is converted so that the sampling frequency becomes a Nyquist rate. It is characterized by comprising first and second downsamplers for downsampling and outputting.

Further, in the above-mentioned audio / musical sound band division encoding / decoding apparatus, the transform coefficient computing means on the encoding side or the decoding side is a fast Fourier transform computing unit.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail with reference to the illustrated embodiments.

FIGS. 1 (a) and 1 (b) are block diagrams showing an encoding device and a decoding device, respectively, used in a speech / music sound band division encoding device according to the present invention.

In the coding apparatus shown in FIG. 1A, reference numerals 1 and 2 denote a low-pass filter and a high-pass filter which divide a band of an input digital signal representing voice and musical sound into a low-frequency band and a high-frequency band, respectively. Is a down-sampler that down-samples the digital signal filtered by each filter at a 2: 1 ratio so that the sampling frequency becomes Nyquist rate.
Reference numerals 5 and 6 denote buffers for storing the down-sampled digital signals in one block (N units), and reference numerals 7 and 8 denote discrete cosine transform of the stored digital signals in blocks.
(DCT), an ATC encoder for adaptive transform coding, 9 is an FFT operator for computing the DCT of these ATC encoders 7, 8 by fast Fourier transform, and 10 is an FFT operator 9
Is divided into several spectral bands, an average power for each band is obtained, and adaptive control of the quantization width in the ATC encoders 7 and 8 is performed based on the auxiliary information obtained from the average power. A bit allocation calculator for allocating an optimal number of quantization bits to each transform coefficient,
Reference numeral 11 denotes a multiplexer that combines the low-frequency and high-frequency data signals quantized and encoded by the ATC encoders 7 and 8 with the auxiliary information and sends the resultant to a transmission channel.

In the decoding device shown in FIG.
Reference numeral 4 denotes a demultiplexer for dividing the data signal from the transmission channel into a low-frequency band and a high-frequency band. Reference numerals 15 and 16 denote each data signal divided by this demultiplexer by inverse discrete cosine transform (inverse DCT) and adaptive transform decoding. ATC decoder 17 performs inverse D of these ATC decoders 15 and 16.
An FFT calculator for calculating a CT by a fast Fourier transform; 18 controls decoding in the ATC decoders 15 and 16 according to a quantization width based on auxiliary information included in the data signal and the number of quantization bits of each transform coefficient. Arithmetic units 19 and 20, buffers for storing each decoded data signal one block at a time, 21 and 22 upsamplers for upsampling the data signal of each buffer at 1: 2, 23 and 24 upsampling A low-pass filter and a high-pass filter for respectively filtering the obtained data signals, and an adder 25 combines the signals passed through both filters and outputs the combined signal as a reproduced signal.

The encoding and decoding device shown in FIG.
It is intended for audio and musical sound signals of about 7 kHz, which are sufficient for high and low band division. The low-pass filters 1 and 23 and the high-pass filters 2 and 24 have a QMF (Quadrature MDF) that does not generate aliasing distortion unless quantization is performed in analysis and synthesis.
irror Filter), and a polyphase structure (poly
phase structure) ("Apprication of q
uadrature mirror filters to split band voice schem
es ", Proc.IEEE Int.Conf.Acoust., Speech, Signal Proce
ssing, Hartford, CT, pp 191 ~ 195: 1977).

Using the coding apparatus having the above configuration (see FIG. 1A), band division coding of voice and musical sound is performed as follows.

A voice / sound signal sampled at a sampling frequency of 16 kHz is input as a digital signal, is divided into two bands by a low-pass filter 1 and a high-pass filter 2, and each of the divided digital signals is The down-samplers 3 and 4 down-sample 2: 1 to 8 kHz sampling signals. Each sampling signal has one block of 32 msec (N = 256)
Are stored in the buffers 5 and 6, and are subsequently adaptively transformed and encoded by the ATC encoders 7 and 8. That is, the FFT calculator 9 calculates a transform coefficient by performing a discrete cosine transform on the sampled signals from the ATC encoders 7 and 8 by FFT, and the bit allocation calculator 10 calculates the transform coefficient as some spectrum. The average power for each band is determined for each band, and based on the auxiliary information determined from the average power, the ATC
Adaptive control of the quantization width in the encoders 7 and 8 and allocation of the optimal number of quantization bits to each transform coefficient are performed. The low- and high-frequency data signals quantized and encoded by the ATC encoders 7 and 8 and the auxiliary information are combined by the multiplexer 11 and transmitted to the transmission channel.

In the above-mentioned band division coding, the band of the input digital signal is divided into a low band and a high band, and each divided digital signal is down-sampled by 2: 1 and thereafter, a single and shared FFT calculator is used. Since the DCT operation is performed in step 9,
The number of complex operation points in one block is halved from 2N = 512 in the case of 16 kHz sampling to N = 256, and therefore, the complex operation amount of one block (see equation (1)) is also 2 × (N / N) from Nlog ₂ N. 2) The amount of operation can be reduced to log ₂ (N / 2), in other words, the amount of operation can be reduced by N (= 256), so that the operation time can be reduced or the scale of the FFT operation unit and, consequently, the scale of the encoding device can be reduced. .

Next, decoding of the signal transmitted to the transmission channel by the above-mentioned encoding device is performed by the decoding device having the above-described configuration shown in FIG. This decoding is performed in the same manner as the above-described coding in the reverse order. In this case, a single FFT calculator is also used for the low-frequency / high-frequency ATC decoder 15,
16, the complex operation amount can be reduced by N, and the operation time or the device scale can be reduced. This effect is particularly effective when the time of one block is increased and N is increased. It becomes remarkable.

In the above embodiment, the case where the band of the input digital signal is divided into two bands, high and low, has been described. However, this band can be divided into several bands by combining 2: 1 band division. .

[0023]

As is apparent from the above description, the band division encoding / decoding apparatus of the present invention divides a digital signal of voice and musical sound into a plurality of bands by a group of band filters, and Since it is possible to perform encoding and decoding by optimally allocating bits according to the situation in consideration of the characteristics of the digital signal comprehensively, it is possible to improve the sound quality of voice and musical sound and to be shared by each divided band. Since a single transform coefficient calculating means can calculate transform coefficients for adaptive transform coding and adaptive transform decoding of each band, it is possible to realize a higher-speed and smaller-scale device than before.

Also, there are first and second downsamplers for downsampling each digital signal output from the first and second bandpass filters on the encoding side so that the sampling frequency becomes a Nyquist rate, and outputting the same. As a result, even if a single transform coefficient calculating means on the encoding side has a small capacity, it is possible to calculate transform coefficients for adaptive transform coding at high speed.

Further, since the transform coefficient computing means on the encoding side or the decoding side is a fast Fourier transform computing unit, the computation of transform coefficients for adaptive transform encoding is performed at a higher speed and on a smaller scale. be able to.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a speech / music sound band division encoding / decoding apparatus according to the present invention.

FIG. 2 is a block diagram showing a conventional SBC device.

FIG. 3 is a block diagram showing a conventional ATC device.

[Explanation of symbols]

 1,23 Low-pass filter 2,24 High-pass filter 3,4 Downsampler 5,6,19,20 Buffer 7,8 ATC encoder 9,17 FFT operator 10,18 Bit allocation operator 11,47,66 Multiplexers 14, 48, 67 Demultiplexers 15, 16 ATC decoders 21, 22, 51, 52 Upsamplers 25, 55 Adders

Claims (3)

(57) [Claims] An encoding side converts an input digital signal of voice / musical sound into at least first and second input digital signals.
The first and second band signals are divided into digital signals of two bands.
Filter and adaptive conversion encoding of the digital signal of the first band.
First adaptive transform coding means, and adaptive transform coding of the digital signal of the second band.
The second adaptive transform coding means and at least the digital signals of the first and second bands.
Used for both the first and second adaptive transform coding means.
A single transform coefficient calculating means for calculating a transform coefficient for at least the digital signals of the first and second bands.
And the characteristics of the digital signals in the first and second bands.
In consideration of the nature, the first and second bits are allocated by performing bit allocation.
Single bit allocation to output to two adaptive transform coding means
Operation means, and the decoding side converts the transmitted data signal into at least the first and second data signals.
Separation means for dividing the data signal of the first band into a first band for adaptive conversion decoding.
Adaptive decoding means for adaptively converting and decoding the data signal of the second band.
And adaptive decoding means shared by at least the data signals of the first and second bands.
For both the first and second adaptive transform decoding means.
A single transform coefficient calculating means for calculating a transform coefficient , shared by at least the first and second band data signals;
And consider the characteristics of the data signals in the first and second bands.
And perform bit allocation to perform the first and second adaptations.
A single bit assignment operator to output to the transform decoding means
Band division of voice and musical sound, characterized by comprising
Encoding / decoding device. 2. The band division of voice and musical sound according to claim 1.
In an encoding / decoding device, first and second bands on the encoding side
Each digital signal output from the bandpass filter
Pull frequency down to Nike straight
First and second downsamplers for sampling and outputting
Band division coding of voice and musical sound characterized by having
Decryption device. 3. The voice / voice according to claim 1 or claim 2.
In a tone band division encoding / decoding device,
Or the transform coefficient calculating means on the decoding side is a fast Fourier transform.
Bandwidth of voice and musical tone characterized by being a conversion arithmetic unit
Split encoding / decoding device.