JPS646574B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field) The present invention relates to an adaptive quantizer that has excellent quantization characteristics even for time-series signals having different statistical properties.

(Background Art) FIG. 1 shows a conceptual diagram of an adaptive quantizer that has been commonly used in the past. In Fig. 1, the sample value e _k of the input signal input as a discrete time series from terminal 1 is multiplied by 1/x _k in multiplier 2 to become z _k , which is encoded into I _o by quantizer 3. . On the other hand, the inverse quantization value e^ _k for the input e _k is calculated by x _k and the inverse quantizer 4
It is obtained by multiplying the inverse quantization value z^ _{k of z^k} by the multiplier 5.

As shown in equation (1), x _k is the sign of the previous sample
Coefficient value M of coefficient table 6 determined by I _o-1
(I _o-1 ) and x _k-1 from one sample before the power multiplier 7
It is obtained by multiplying x ^B _k-1 raised to the B power by the multiplier 8.

x _k = M (I _o-1 ) x ^B _k-1 (1) Conventionally, the adaptive quantizer according to the above algorithm requires a coefficient table 6 dedicated only to the audio signal. It had good quantization properties. However, this method has the disadvantage that quantization characteristics deteriorate for time-series signals with different statistical properties.

(Problem to be solved by the invention) The present invention provides an adaptive quantum system that does not degrade quantization characteristics even for time-series signals with different statistical properties, especially audio signals and voice band signals (voice band data) with different frequency spectrum structures. In order to configure the quantizer, at least two types of coefficient tables are provided depending on the input signal, and the tables are switched based on the result determined from the inverse quantization output signal.

(Structure and operation of the invention) FIG. 2 shows a first embodiment of the invention. Portions 11 and 23 surrounded by dotted lines are circuits added to the conventional circuit shown in FIG. Extract the features of the input signal and select 6 or 10 depending on the result.
This is a circuit that outputs a signal S for switching the coefficient table.

FIG. 3 shows a second embodiment of the invention. In the figure, the center frequency of the dequantized value e^ _k is the first
The signals are input to band filters 15 and 16 near the formant frequency and the carrier frequency of the voice band data, and the powers of the output signals of the two filters are successively calculated by accumulators 17 and 18. Then, a comparator 19 successively obtains the ratio α of the two powers, and a successive determination circuit 20 determines the range of the value, and successively outputs a switching signal.

FIG. 4 shows a third embodiment of the invention. 20 is a sequential determination circuit, and 21 is a buffer circuit.

If the signal α contains noise, the sequential judgment circuit 2
0 may be affected by noise and may malfunction. The buffer circuit 21 removes the influence of noise.
A preset type counter is used to issue a table switching command only when a preset condition is met. When the circuit of FIG. 4 is added to FIG. 3, it is assumed that whether or not the circuit of FIG. 4 is operated and the number preset in the preset type counter can be arbitrarily set.

There are two types of coefficient tables suitable for audio signals and voice band data, which have different statistical properties, and these tables can be switched appropriately depending on the input signal. has the advantage of being able to adaptively quantize signals sent in time series.

(Effects of the Invention) The present invention has quantization characteristics that are highly adaptable to signals with different statistical properties.
It can be used as a quantizer when image signals and voice band data are sent in time series.

[Brief explanation of drawings]

Figure 1 is a conceptual diagram of a conventional adaptive quantizer; Figure 2 is a conceptual diagram of a conventional adaptive quantizer;
The figure is a conceptual diagram showing a first embodiment of the present invention, FIG. 3 is a conceptual diagram showing a second embodiment of the present invention, and FIG. 4 is a conceptual diagram showing a third embodiment of the present invention. 1, 12, 13, 14... terminal, 2, 5, 8...
...multiplier, 3...quantizer, 4...inverse quantizer,
6, 10... Coefficient table, 11... Judgment circuit,
15, 16... band filter, 17, 18... accumulator, 19... comparator, 20... sequential determination circuit,
21... Buffer circuit.

Claims (1)

[Claims] 1. In a Gudman-type adaptive quantizer that sequentially quantizes a discrete input signal sequence according to the contents of a coefficient table that defines a quantization width, the frequency spectrum input as the input signal sequence is At least two types of coefficient tables are provided corresponding to audio signals and audio band signals with different structures, and at least two feature extraction filters with different bandwidths that receive output signals of inverse quantization are provided, and the feature extraction An adaptation characterized in that it is provided with a determining means for determining whether the signal is an audio signal or an audio band signal based on the output signal of the filter and outputting a switching signal, and a switching means for switching the coefficient cable based on the switching signal. Shape quantizer. 2. The adaptive quantizer according to claim 1, wherein the determining means determines the power ratio of each output signal of the feature extraction filter, and performs the determination according to a range of values of the power ratio. 3. A patent claim characterized in that it is possible to arbitrarily set whether the determining means outputs a switching signal based on the result of the determination or outputs the switching signal when the result of the determination continues a predetermined number of times. range 2nd
Adaptive quantizer as described in section.