JPH04129430A - Band split coder - Google Patents

Abstract

PURPOSE:To attain optimum coding even to any voice signal by obtaining power for each band and quantizing the power to 2<2sigma> when a band of the digital voice signal is split. CONSTITUTION:The coder is provided with a band split filter splitting a digital signal such as a voice signal into plural bands, a power calculation means 2 collecting voice signals split into bands to be blocks for several samples each and calculating the power in each block, a quantizing means 3 quantizing the power to 2<2sigma>, a quantization means 6 quantizing the power adaptively for each block and a quantization bit control section 5 controlling adaptively the quantization bit for each band depending on the value sigma. Thus, a bit number assigned to each band is obtained adaptively in response to an input signal. Thus, the bit number assigned to each band with a simple processing only is made adaptive so as to be most efficient to the signal at that time.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application The present invention relates to a band division encoding device for encoding digital audio signals used in audio IC recorders and the like.

(b) Prior Art Conventionally, when compressing and encoding a digital audio signal, there is a method of dividing it into several subbands and encoding them.

This is an example of IBM's 5pilit-Band
Voice Coding Scheme (IEEE
CASSP 1977 MAY).

The input digital audio signal is first divided by a group of band-pass filters (QMF filters), and each of the divided digital audio signals is frequency-shifted down by the center frequency of the passband, and then passed through the band-pass filter. Downsample at the sampling frequency determined by the bandwidth,
The results are encoded using a forward adaptive PCM code for each block.

The decoding section performs the reverse operation for each band to create a band-pass signal for each band, and adds them together to form an output audio waveform.

The advantages of this method are (1) It is easy to effectively allocate bits for each band.

(2) The quantization error in each band is related only to the signal in that band, and if the power in that band is small, even if there is a lot of quantization distortion, it is only a small error when viewed from the overall signal.

(3) Quantization errors in each band ideally become white noise that spreads widely in terms of frequency spectrum, with only a portion of it falling within that band.

Examples include.

Also, from (1) and (2) above, it is clear that it is desirable to change the number of bits per sample allocated to each band depending on the power of the signal in that band, but this generally results in complicated processing. Therefore, it is common to allocate a fixed number of bits, such as 4.2°2.2 in the case of 4 bands, with more in the low range and less in the high range.

In some cases, it may be changed adaptively, but in that case, it is considered most appropriate to divide according to the following formula, and this calculation requires a very complicated device.

U, = U, /W where R is the average bit rate per sample R, is i
The number of bits U to be allocated per sample of the band, 1. ti band power W+1. Bandwidth ratio of tt band (c) Problems to be solved by the invention The present invention adapts the number of bits allocated to each band to be most efficient for the signal at that time in a band division encoding device by simple processing. The present invention provides a highly efficient encoding device.

(d) Means for Solving the Problems The present invention is intended to solve the above problems, and provides a band division filter that divides a digital signal such as an audio signal into a plurality of bands, and a block-by-block output of the filter. A power calculation means that calculates the power of The present invention provides a band division encoding device comprising a section.

(e) Effect: By the means described above, the number of bits to be allocated to each band can be adaptively determined according to the input signal.

(F) Embodiment Next, an embodiment of the present invention will be explained using the block diagram of FIG. (1) is a QMF filter that divides the input audio into high and low frequencies at a frequency of 1/4 of the sampling frequency and downsamples it to 1/2 of the sampling frequency.In this example, the QMF filter (1) By using two stages of filters, the input audio is divided into four bands and the output of the filter (1) is reduced to the baseband. (2
) is a power calculation means that collects audio divided into bands for each band into blocks every few samples and calculates the power within the block. (3) is a quantization means that quantizes the calculated power to 2″′, which actually represents the power as a binary number, searches for the leftmost 1, and determines the number of bits from the right. Find σ1 for each band by subtracting 1 from that value and dividing by 2. (4) is a normalization method that normalizes the power within one block to 1, and has a wide dynamic range. It is now possible to clearly quantize digital audio signals, and in reality, each band's signal is converted to the square root of the quantized power of that band (that is, 2
It can be found by dividing by ゛). (5) is a quantization allocation means for allocating quantization bits to be used in a quantizer for each band, which will be described later, and can be allocated according to the following equation using σ1 of each band.

1-NWl=1/2-10g1 (WITTW,
−”I) j=1 If W+=1/4, then R+ = R+σ1− (σ, +σ, +σ, +σ,)/
4 Here, R1 is the number of quantization bits allocated to one sample in the l band, and R is the number of quantization bits allocated to one sample, averaged over the entire range. Furthermore, since R1 may actually be found as a decimal or negative number, a simple conditional statement is required to convert these numbers into positive integers.

(6) is a quantization means for quantizing the signal normalized by the normalization means (4) according to the number of quantization bits allocated by the quantization allocation means (5), and (7) is a quantization means for each band. This is a multiplexer that multiplies the encoded code and auxiliary information σ. (8) is a demultiplexer that divides the signal from the multiplexer (7) into codes of each band and auxiliary information σ. (9) is a decoding allocation means that calculates bit allocation in the same way as the quantization allocation means (5);
10) is a decoding means that decodes the quantized code of each band, (11) is a power decoding means that decodes the power of each band, and (12) is a normalized code using the decoded power. The restoring means (13) for restoring the converted signal to its original state is a QMF filter on the restoring side, which performs high frequency conversion and band combining of the band-divided and down-sampled signal.

For any voice, the filter (1) performs band division to obtain the power for each band, and the quantization means (6) quantizes the value to 21', thereby dividing the power based on the σ value. ,
Since the quantization bits assigned to each band can be adaptively controlled by the quantization assignment means (5), optimal encoding can be performed. By further retracing the previous operations, the encoded signal can be decoded into the original digital audio signal.

(G) Effects of the Invention According to the device of the present invention, when dividing a digital audio signal into bands, the power for each band is calculated and the value is 2! This quantization makes it possible to adaptively control the quantization bits assigned to each band, which is extremely useful because any voice can be optimally encoded.

[Brief explanation of drawings]

FIG. 1 is a block diagram for explaining one embodiment of the apparatus of the present invention. (1)... Bandwidth division (QMF) filter, (2)...
- Power calculation means, (5)...Quantization bit control unit (quantization bit allocation means), (6)...Quantization means.

Claims (1)

[Claims] (1) A band division filter that divides a digital signal such as an audio signal into a plurality of bands, and a power calculation means that calculates the power of each block of the output of the filter and quantizes it in the form of 2^2;
A band division encoding device comprising: a quantization unit that performs adaptive quantization for each block; and a quantization bit control unit that adaptively controls quantization bits for each band based on the value of σ.