JPH04123535A - Band split type voice coding device - Google Patents

Abstract

PURPOSE:To realize the voice coding device in which quantization bit allocation number is not zero even when power is concentrated on low frequencies or voice signal power is in existence even at high frequencies by providing a specific high frequency bit allocation discrimination section and a bit re-assignment section respectively in the voice coding device. CONSTITUTION:A high frequency bit allocation discrimination section 2 discriminates whether or not a bit, number allocated at high frequencies of a high band is 0 as to a quantization bit number obtained and allocated by a bit allocation section 1 in terms of a coding bit rate and a bit allocation number is used as a quantity bit allocation number as it is when not 0. When 0, a bit re-allocation section 3 applies bit re-allocation at a bit rate higher than the coding bit rate in which a bit number used for quantization of one frame. Thus, when voice signal power exists even in high frequencies, the bit allocation number is not 0 in the high frequencies and in this case, one bit is allocated, for example, and remaining bits are allocated to other frequencies. Thus, even when power is concentrated on low frequencies or voice signal power is in existence in also high frequencies, the quantization bit number is not 0 and no unclear voice is caused.

Description

[Detailed Description of the Invention] [Summary] In the bit allocation unit, the average number of bit allocation is calculated by dividing the number of bits that can be used for quantizing the residual signal of one frame of the audio signal by the number of samples of one frame, and The obtained average bit allocation number is allocated to each frequency band according to the magnitude of the predicted residual signal power in each band where the input audio signal is divided into multiple frequency bands, and in each frequency band, the allocated bit number is Regarding a band division type audio encoding device that performs quantization by number, even if the power is concentrated in the low frequency range, if there is audio signal power in the high frequency range, the band division type audio coding device that does not set the number of quantization bits allocated to O is used. For the purpose of providing a speech encoding device, on the output side of the bit allocation unit, it is determined whether the number of bits allocated to the high frequency band in the bit allocation unit is O, and if it is not O, the number of bits is determined to be O. A high-frequency bit allocation determination unit that uses the actual bit allocation number as the bit allocation number for quantization, and if it is O, it is processed by the bit re-allocation unit; and if it is 0, it is re-allocated at a bit rate higher than the encoding bit rate. The configuration includes a refocusing unit that performs the refocus assignment.

[Industrial Application Field] The present invention calculates the average bit allocation number by dividing the number of bits that can be used for quantizing the residual signal of one frame of an audio signal by the number of samples of one frame in a bit allocation unit, The obtained average focus allocation number is divided into multiple frequency bands according to the magnitude of the predicted residual signal power of each band.
The present invention relates to an improvement in a band division type speech encoding device that allocates each frequency band and performs quantization using the allocated number of bits in each frequency band.

[Prior Art] Fig. 5 is a block diagram of a conventional band division type speech encoding device, Fig. 6 is a flowchart of the processing of the 8 Kbps bit allocation unit of the bit allocation unit in Fig. 5, and Fig. 7 is a block diagram of a conventional band division type speech encoding device. FIG. 3 is a diagram showing an example of spectrum analysis when a signal is subjected to 8Kbps encoding processing.

In FIG. 5, the audio band is divided into three parts, high, middle, and low, by a band division filter 20.21, and predictive encoders 22, 23, and 24 are used to generate predictive codes for each high, middle, and low band. The resulting residual code is multiplexed by a multiplexer 25 and transmitted as a transmission code.

This predictive coding for each high, middle, and low band is performed using a fixed frame length, and is quantized using the number of bits allocated to subintervals obtained by dividing the pitch period. Bit allocation will be explained by taking as an example the case where the number of partial sections divided into four is 4.

In the prediction residual signal calculation unit 11 of the bit allocation unit 1, high frequency,
The prediction residual signals of the middle and low bands are calculated, passed to the prediction residual signal power calculation section 12, and the prediction residual signal calculation section output section 12,
Calculate the power of the predicted residual signal for high, middle, and low frequencies, and
The data is passed to the bps bit allocation unit 13.

In the 8Kbps bit allocation unit 13, in step 1 of FIG. The number of bits divided by the number of samples in one frame) Rcm is calculated by dividing the number of bits by the number of samples in one frame.
The bit allocation number is determined by allocating to the high, middle, and low bands on the frequency axis of 8 Kbps according to the predicted residual signal power of the low band.

Then, in step 2, find the high range and mid range.

The number of bits allocated to the low frequency band is set as the number of bits allocated to each subinterval according to the subinterval residual signal power of each of the four subintervals, and the number of bits allocated is corrected by the correction number Drl so that the allocated number becomes an integer. The number of bits allocated on the time axis is 8 Kbps, and quantization is performed by the quantizers of the predictive encoders 22, 23 and 24.

[Problem to be solved by the invention]

However, because the number of quantization focuses is assigned according to the predicted residual signal power of the high, middle, and low frequencies, the power of the audio signal is generally concentrated in the low range; is concentrated, the number of bits allocated may become 0 for all subintervals, regardless of the presence of audio signal power in the high frequency range.

In such a case, the spectrum of the input audio signal shown in FIG. 7(A) and the spectrum of this audio signal shown in FIG.
As can be seen by comparing the spectrum with the bps encoding process, there is a problem in that the voice is muffled due to the lack of high frequency components.

An object of the present invention is to provide a band division type audio encoding device in which the number of quantization bits allocated is not set to O even if the power is concentrated in the low frequency range but there is also audio signal power in the high frequency range.

[Means to solve the problem]

FIG. 1 is a block diagram of the principle of the present invention.

As shown in FIG. 1, in the bit allocation section 1, one
The number of bits that can be used for quantizing the residual signal of a frame is set to 1.
The average number of bit allocations is calculated by dividing the number of samples in the frame, and the calculated average number of bit allocations is divided into multiple frequency bands.The input audio signal is divided into multiple frequency bands. In a band division type audio encoding device that allocates to each frequency band and performs quantization using the allocated number of bits in each frequency band, the bit allocation unit 1
On the output side of the bit allocation unit 1, it is determined whether the number of bits allocated to the high frequency band is O, and if it is not 0, the bit allocation is set as the quantization bit allocation number, and If it is 0, there is provided a high-frequency bit allocation determination unit 2 that causes the bit re-allocation unit 3 to perform the process, and if it is 0, a bit re-allocation unit 3 that performs bit re-allocation at a bit rate higher than the encoding bit rate.

[For production]

According to the present invention, for the number of quantization bits determined and allocated by the bit allocation unit 1 based on the encoding bit rate, the high-frequency bit allocation determination unit 2 determines that the number of bits allocated to the high frequency band is O. Determine whether or not, and if it is not 0, then the value of Bi-7
The number of bits to be allocated is set as the number of quantization bits to be allocated, and if it is 0, the bit re-allocation unit 3 re-allocates bits at a bit rate higher than the encoding bit rate, which increases the number of bits that can be used for quantization of one frame. Therefore, if there is audio signal power in the high frequency range as well, the number of bits allocated to the high frequency range will not be 0. In this case, for example, 1 bit will be allocated and the rest will be allocated to other areas.

Therefore, even if the power is concentrated in the low frequency range, if there is also audio signal power in the high frequency range, the number of quantization bits allocated will not be 0, and the voice will not sound muffled.

〔Example〕

FIG. 2 is a block diagram of a band division type speech encoding device according to an embodiment of the present invention, and FIG. 3 is a block diagram of the 8Kbp bit allocation section of FIG.
FIG. 4 is a flowchart of the processing after the s bit allocation section, and is a diagram showing an example of quantization bit allocation in the high frequency range.

The difference in FIG. 2 from the conventional example in FIG. 5 is that the bit allocation unit 1
0, the bit allocation unit 1 and the high frequency bit allocation determination unit 2
Since this is the addition of the bit re-allocation section 3, this different point will be explained below with reference to FIG.

The steps up to step 2 in FIG. 3 are the same as up to step 2 in FIG.
I will explain from.

In step 3 of FIG. 3, the high-frequency bit allocation determination unit 2 in FIG. If the bit allocation is 0, the process proceeds to the 16 Kbps bit allocation unit 14 of the bit re-allocation unit 3 in FIG. 2, and in step 4 of FIG. Average number of bits allocated per sample RCI&
, high range.

Distributed according to the residual signal power of the mid-range and low-range, 16Kbp
The number of bits allocated to the high, middle, and low ranges on the frequency axis of s is the number of bits allocated, and then the high range obtained in step 5.

The number of allocated bits for the mid-range and low-range is distributed to each of the four sub-intervals according to the residual signal power of the sub-interval, and the number of allocated bits is corrected using a correction number DrI6 so that it becomes an integer. ,1
As the focus allocation number on the time axis of 6 Kbps, the process proceeds to the high frequency bit allocation determination unit 15 in FIG.

In step 6 of FIG. 3, the high frequency bit allocation determining unit 15 determines whether the bit allocation for all subintervals of the high frequency band is 0.
If so, set the number of bits allocated to the four subintervals of high, middle, and low ranges obtained in step 2 as the number of bits allocated, and if it is not 0, use the high-range bit allocation section of FIG. 2. 16, and in step 7 of FIG. 3, a subsection with a large inner frame difference signal power of each subsection of the high frequency band is extracted. For example, if subsection 1 is large, the subsections shown in FIGS. 7(A) and (C) As shown in FIG.
If is large, as shown in FIG. 7(B), partial interval 1,
2, and proceeds to the middle/low frequency range bit re-allocation section 17L in FIG.

In step 8, the mid-low band re-allocation unit 17 calculates the focus allocation on the 8 Kbps time axis for the mid-range and low band from the remainder after subtracting the bits allocated to the high band, and calculates the focus allocation on the 8Kbps time axis as the number of bit allocation. Assigned to 22°23.

In this way, even if the audio power is concentrated in the low range,
If there is audio signal power in the high frequency range, the number of bits allocated to the high frequency range will never become 0, so the voice will not sound muffled.

(Effects of the Invention) As explained in detail above, according to the present invention, even if audio power is concentrated in the low frequency range, if there is audio signal power in the high frequency range, the number of bits allocated to the high frequency range is O. Since it will not be
This has the effect of preventing your voice from sounding muffled.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of the principle of the present invention. FIG. 2 is a block diagram of a band division type speech encoding device according to an embodiment of the present invention. FIG. Processing flowchart, Figure 4 is a diagram showing an example of quantization bit allocation in the high frequency range, Figure 5 is a block diagram of a conventional band division type audio encoding device, and Figure 6 is the bit allocation shown in Figure 5. FIG. 7 is a diagram showing an example of spectrum analysis when an input audio signal is subjected to 8Kbps encoding processing. In the figure, 1.10 is a bit allocation unit, 2.15 is a high-frequency bit allocation determination unit, 3 is a bit re-allocation unit, 11 is a predicted residual signal calculation unit, 12 is a predicted residual signal power calculation unit, and 13 is a predicted residual signal power calculation unit. 8 Kbps bit allocation section; 14 is a 16 Kbps bit allocation section; 16 is a high-frequency bit allocation section; 20.21 is a band division filter; 22.23.24 is a predictive encoder; 25 is a multiplexing section. =O Rhinoceros f! of the present invention! Brosoff's diagram 1 Diagram Part Section Number Part Section 1 Quantization in Honjo J2 1, 1-11 Closed square diagram showing the relevant minister

Claims (1)

[Claims] In the bit allocation unit (1), the average number of bits allocated is calculated by dividing the number of bits that can be used for quantizing the residual signal of one frame of the audio signal by the number of samples of one frame. The average number of bits is allocated to each frequency band according to the magnitude of the predicted residual signal power in each band where the input audio signal is divided into multiple frequency bands. In a band division type speech encoding device that performs quantization, the bit allocation unit (
On the output side of 1), the bit allocation unit (1) determines whether the number of bits allocated to the high frequency band is 0 or not, and if it is not 0, the bit allocation part (1) determines whether the number of bits allocated to the high frequency band is 0 or not. If the bit allocation number is 0, the high-frequency bit allocation determination unit (2) causes the bit re-allocation unit (3) to process the bit allocation, and if it is 0, the bit allocation is performed at a higher bit rate than the encoding bit rate. What is claimed is: 1. A band division type speech encoding device comprising: a bit re-allocation section (3).