JPH0736493A - Variable rate speech coder - Google Patents

Abstract

(57) [Summary] [Purpose] The order of the linear analysis filter is made variable for each frame so that the quantization bit of the linear prediction coefficient can be easily changed. [Structure] A frame cutout unit 11 cuts out one frame of voice data from input voice, and a linear prediction coefficient calculation unit 12 calculates a PACOR coefficient in the one frame of voice data. Further, the order determining unit 13 determines the order p of the linear prediction analysis. The linear prediction coefficient calculation unit 12 sends the linear prediction coefficients α1 to αp corresponding to the order p to the linear analysis filter 14 and the quantization unit 15. The linear analysis filter 14 calculates the residual signal, and the quantizer 15 quantizes the linear prediction coefficient into a coded code and sends it out. As a result, the order p of the linear prediction coefficient becomes variable for each frame, and the processing at the variable rate of the linear prediction coefficient is performed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a variable rate speech coding apparatus which is used in a digital speech communication apparatus, a digital speech storage apparatus or the like and which transmits coded data at a variable rate.

[0002]

2. Description of the Related Art In recent years, research on variable-rate speech coding processing has been conducted as speech coding. This variable-rate voice encoding process differs from the conventional fixed-rate voice encoding process in which the amount of transmission information is fixed, and the amount of transmission information changes with time. In this case, it is possible to reduce the total amount of transmission information as compared with the fixed rate speech coding process. Further, in both the variable rate speech coding processing and the fixed rate speech coding processing, the order of the linear prediction analysis of speech is mainly fixed.

FIG. 3 is a block diagram showing the structure of a conventional variable rate speech coding apparatus. In FIG. 3, the variable rate audio encoding device includes a frame cutout unit 6 that cuts out audio data of one frame length from input audio (signal).
And a linear prediction coefficient calculation unit 7 that calculates a linear prediction coefficient,
It has a linear analysis filter 8 for calculating a residual signal and a quantizer 9 for quantizing a linear prediction coefficient into an encoding code. The order of the linear analysis filter 8 here is a fixed value.

Next, the operation of this conventional configuration will be described. The frame cutout unit 6 cuts out audio data of one frame length. The linear prediction coefficient calculator 7 calculates the linear prediction coefficients α1 to αp such that the output power of the linear analysis filter 8 is minimized. The linear analysis filter 8 calculates the residual signal. Further, the quantizing unit 9 quantizes the linear prediction coefficients α1 to αp into coding codes.

[0005]

In such a conventional speech coding process as described above, since the order p of the linear analysis filter 8 is fixed, the number of bits of the quantization code of the linear prediction coefficients α1 to αp is made variable. It had a problem that it was difficult to do.

The present invention solves such a conventional problem, and makes it possible to easily change the quantization bit of the linear prediction coefficient by changing the order of the linear analysis filter for each frame. An object is to provide a speech encoding device.

[0007]

In order to achieve the above object, a variable rate speech coding apparatus according to the present invention is provided with a frame clipping means for clipping speech data of one frame from input speech and speech data of one frame. , A linear predictive coefficient calculating means for calculating the coefficient, an order determining means for determining the order of the linear predictive analysis, a linear predictive coefficient calculating means for sending a linear predictive coefficient corresponding to the order, and a variable order for calculating the residual signal. , And a quantizing means for quantizing and transmitting the linear prediction coefficient into a coded code. By varying the order of the linear prediction coefficient for each frame, processing is performed at a variable rate of the linear prediction coefficient. It is configured.

In addition, the linear prediction coefficient calculation means uses PACOR.
The coefficient is calculated, and the linear analysis filter uses the linear prediction coefficient for the order to calculate the prediction gain indicating the output power to input power ratio.

[0009]

With such a configuration, the variable rate speech coding apparatus of the present invention has the order of the linear prediction coefficient for each frame,
That is, the numerical value of the linear prediction coefficient becomes variable, and the number of bits of the coding code in the quantizing unit of the linear prediction coefficient is easily variable.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of a variable rate speech coder according to the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of a variable rate speech coding apparatus of the present invention. In FIG. 1, this variable rate speech coding apparatus has a frame cutout unit 11 that cuts out one frame of speech data of an input speech (signal), and a linear prediction coefficient calculation unit 12 that calculates a PACOR coefficient. . Further, this variable rate speech encoding apparatus has an order determining unit 1 for determining the order p.
3, a variable-order linear analysis filter 14 for calculating a residual signal, and a quantizer 15 for quantizing a linear prediction coefficient into an encoding code.

Next, the operation of the configuration of this embodiment will be described. The frame cutout unit 11 cuts out one frame of sound data from the input sound that is continuous sound, and the linear prediction coefficient calculation unit 12 calculates the PACOR coefficients k1 to kpmax in the cutout one frame of sound data. Here, pmax is the maximum order.
Then, the order determining unit 13 executes the order p through the following processing procedure.
To decide. (1) PACOR coefficient k at each order p from input speech
i (1 ≦ i ≦ pmax) is calculated. (2) The prediction gain ui (0 ≦ i ≦ pmax) of the linear analysis filter 14 at each order p is calculated by the following equation [Equation 1]. The prediction gain ui indicates the output power to input power ratio in the linear analysis filter 14 using the linear prediction coefficient for the order i.

[0013]

[Equation 1]

(3) The order p is determined using the following equation (2). The order p is the minimum value of i satisfying ui <ut (0 ≦ i ≦
pmax) where ut = upmax × (1 + du / 100) (2) du: parameter that determines the operation of the order determination process (%) (4) The linear prediction coefficient calculation unit 12 corresponds to the order of the order p. The prediction coefficients α1 to αp are calculated. (5) Using the linear prediction coefficients α1 to αp calculated by the linear prediction coefficient calculation unit 12, the linear analysis filter 14 calculates the output residual.

In this case, depending on the input voice, there is a signal that has no coding effect even if the order p of the linear analysis filter 14 is increased. For example, with respect to white noise, the linear predictors are close to 0, and the linear analysis filter 14 has no meaning. This is because the linear analysis filter 14 performs processing for flattening the spectral envelope of the input voice. Therefore, the linear analysis filter 14 cannot be used for a signal having a flat spectrum such as white noise. This means that the order p of the linear analysis filter 14 may be reduced depending on the nature of the input voice.

Therefore, in the above-described linear prediction coefficient order determination process, the order p is calculated when the prediction gain is close to the prediction gain at the maximum order. At this time, the output residual when the linear analysis filter 14 is applied with the order p and the output residual when the linear analysis filter 14 is applied with the maximum order have the same waveform.

FIG. 2 is a waveform diagram showing an example of an actual voice waveform. In FIG. 2, in this example, the waveform when the order p of the input voice shown in FIG. 2 (a) is 2 is shown in FIG. 2 (b). 2 (c), (d), (e), and (f) show waveforms when the orders p are 4, 5, 8, and 10, respectively.

By changing the order p for each frame in this way, the number of linear prediction coefficients to be transmitted in this speech coding process becomes smaller than the number of linear prediction coefficients in the fixed rate speech coding process.

[0019]

As is apparent from the above description, since the variable rate speech coding apparatus of the present invention makes the value of the linear prediction coefficient which is the order of the linear prediction coefficient variable for each frame, the linear prediction coefficient is changed. This has the effect that the number of bits of the encoded code in the quantization unit can be easily changed.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of a variable rate speech coding apparatus of the present invention.

FIG. 2 is a waveform diagram showing an example of an actual voice waveform in the embodiment.

FIG. 3 is a block diagram showing a configuration of a conventional variable rate speech coding apparatus.

[Explanation of symbols]

 11 frame cutout unit 12 linear prediction coefficient calculation unit 13 order determination unit 14 linear analysis filter 15 quantization unit

Claims (2)

[Claims] 1. A frame cutout unit that cuts out one frame of voice data from an input voice, a linear prediction coefficient calculation unit that calculates a coefficient in the one frame of voice data, and an order determination unit that determines the order of linear prediction analysis. Means, a linear prediction coefficient calculation means for sending a linear prediction coefficient corresponding to the above-mentioned order, a variable-order linear analysis filter for calculating a residual signal, and a quantization for quantizing and sending the linear prediction coefficient into a coding code Means for varying the order of the linear prediction coefficient for each frame and performing processing at a variable rate of the linear prediction coefficient. 2. A linear prediction coefficient calculation means calculates a PACOR coefficient, and a linear analysis filter uses a linear prediction coefficient for an order to calculate a prediction gain indicating an output power to input power ratio. The variable rate speech coding apparatus according to claim 1.