Voice coding/decoding system having selected coders and entropy coders

Claims

We claim:

1. A voice coding/decoding system having a transmitting part for transmitting a coded signal of an input voice signal at a bit rate lower than a predetermined transmission bit rate and a receiving part receiving and decoding the coded signal transmission form said transmission part, said transmitting part comprising:

a plurality of coders, each having an output, for receiving an input voice signal, and for coding the input voice signal, said coders having predetermined numbers of quantizer output levels and having a predetermined quantization characteristic;

a plurality of groups of entropy coders, each of said entropy coders in each of said groups having an input operatively connected to the output of one of said plurality of coders and outputs, said entropy coders having predetermined numbers of quantizer output levels and predetermined probability distributions;

evaluation means, operatively connected to the outputs of said plurality of coders and the outputs of said plurality of entropy coders, for evaluating characteristics of the outputs of said coders and said entropy coders in order to select, from said entropy coders, selected entropy coders having output bit rates lower than the transmission bit rate and to select, from the coders connected to the selected entropy coders, a selected coder having the best output characteristic, by outputting a selecting signal indicating a combination of the selected coder and an entropy coder in the selected entropy coders; and

selecting means, operative connected to the outputs of said plurality of entropy coders and to the output of said evaluation means, for selecting, in response to the selecting signal a codeword passed through said combination of the selected coder and entropy coder to be transmitted.

2. A voice coding/decoding system as claimed in claim 1, wherein said predetermined numbers of quantizer output levels of said plurality of coders are the same as each other, said predetermined quantization characteristics of said plurality of coders are different from each other, the numbers of said predetermined quantization output levels of said entropy coders in each of said groups are the same as each other, and the quantization characteristics in each of said groups of entropy coders are different from each other.

3. A voice coding/decoding system as claimed in claim 1, wherein said predetermined numbers of quantizer output levels of said plurality of coders are different from each other, said predetermined quantization characteristics of said plurality of coders are the same as each other, the numbers of said predetermined quantization output levels of said entropy coders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy coders are the same as each other.

4. A voice coding/decoding system as claimed in claim 1, wherein said predetermined numbers of quantizer output levels of said plurality of coders are different from each other, said predetermined quantization characteristics of said plurality of coders are the same as each other, the numbers of said predetermined quantization output levels of said entropy coders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy coders are different from each other.

5. A voice coding/decoding system as claimed in claim 1, wherein said predetermined numbers of quantizer output levels of said plurality of coders are different from each other, said predetermined quantization characteristics of said plurality of coders are different from each other, the numbers of said predetermined quantization output levels of said entropy coders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy coders are different from each other.

6. A voice/decoding system as claimed in claim 1, wherein said selecting means comprises multiplexing means for multiplexing the codeword passed through said combination with the selecting signal.

7. A voice coding/decoding system as claimed in claim 1, wherein at least one of said coders is directly connected to said selecting means without connecting through said entropy coders.

8. A voice coding/decoding system as claimed in claim 1, wherein said plurality of coders comprise adaptive differential pulse code modulation coders.

9. A voice coding/decoding system as claimed in claim 1, wherein each of said plurality of coders includes a quantizer.

10. A voice coding/decoding system as claimed in claim 1, wherein each of said plurality of coders comprises a local decoder for generating a local decoded signal by locally decoding the codeword output from the coder, said local decoded signal being supplied to said evaluation means for extracting a coder having said best characteristic.

11. A voice coding/decoding system as claimed in claim 10, wherein the characteristics of said coders are signal to noise ratios in said local decoded signal.

12. A voice coding/decoding system as claimed in claim 10, wherein the characteristics of said coders are the summation of the absolute values of the error signals in each local decoded signal.

13. A voice coding/decoding system as claimed in claim 10, wherein the characteristics of said coders are the peak value of the absolute value of the error signals in said local decoded signal.

14. A voice coding/decoding system as claimed in claim 10, wherein the characteristics of said coders are the cepstrum distances in said local decoded signal.

15. A voice coding/decoding system as claimed in claim 1, wherein said entropy coders are Huffman coders.

16. A voice coding/decoding system as claimed in claim 1, wherein said entropy coders are arithmetic coders.

17. A voice coding/decoding system as claimed in claim 1, wherein said entropy coders are runlength coders.

18. A voice coding/decoding system as claimed in claim 1, wherein said entropy coders are Ziv-lempel coders.

19. A voice coding/decoding system as claimed in claim 1, wherein said receiving part comprises:

demultiplexing means, operatively connected to the output of said transmitting part through a transmission line, for demultiplexing a received signal from said transmitting part into the codeword and the selecting signal;

switching means, operatively connected to said demultiplexing means, having a single input end for receiving the codeword from said demultiplexing means and a plurality of output ends, for passing said codeword to one of said plurality of output ends in response to the selecting signal from said demultiplexing means;

a plurality of groups of entropy coders, each of said entropy coders in each of said groups having an input operatively connected to said plurality of output ends of said switching means and outputs, said entropy coders in each of said groups having predetermined numbers of quantizer output levels and predetermined probability distributions corresponding to a corresponding group of entropy coders in said transmitting group; and

a plurality of decoders each having an input, the outputs of said entropy decoders in each of said groups being operatively connected to the input of one of said plurality of decoders, for decoding an entropy decoded codeword from said entropy decoders, said decoders having predetermined numbers of quantizer output levels and having the predetermined quantization characteristics;

whereby, in accordance with the selecting signal indicating the best combination of a coder and an entropy coder, by the corresponding combination of one of said entropy decoders and one of said coders, the transmitted codeword is decoded.

20. A voice coding/decoding system as claimed in claim 19, wherein said predetermined numbers of quantizer output levels of said plurality of coders are the same as each other, said predetermined quantization characteristics of said plurality of coders are different from each other, the numbers of said predetermined quantization output levels of said entropy coders in each of said groups are the same as each other, and the quantization characteristics in each of said groups of entropy coders are different from each other.

21. A voice coding/decoding system as claimed in claim 20, wherein said predetermined numbers of quantizer output levels of said plurality of decoders are the same as each other, said predetermined quantization characteristics of said plurality of decoders are different from each other, the numbers of said predetermined quantization output levels of said entropy decoders in each of said groups are the same as each other, and the quantization characteristics in each of said groups of entropy decoders are different from each other.

22. A voice coding/decoding system as claimed in claim 19, wherein said predetermined numbers of quantizer output levels of said plurality of coders are different from each other, said predetermined quantization characteristics of said plurality of coders are the same as each other, the numbers of said predetermined quantization output levels of said entropy coders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy coders are the same as each other.

23. A voice coding/decoding system as claimed in claim 22, wherein said predetermined numbers of quantizer output levels of said plurality of decoders are different from each other, said predetermined quantization characteristics of said plurality of decoders are the same as each other, the numbers of said predetermined quantization output levels of said entropy decoders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy decoders are the same as each other.

24. A voice coding/decoding system as claimed in claim 19, wherein said predetermined numbers of quantizer output levels of said plurality of coders are different from each other, said predetermined quantization characteristics of said plurality of coders are the same as each other, the numbers of said predetermined quantization output levels of said entropy coders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy coders are different from each other.

25. A voice coding/decoding system as claimed in claim 24, wherein said predetermined numbers of quantizer output levels of said plurality of decoders are different from each other, said predetermined quantization characteristics of said plurality of decoders are the same as each other, the numbers of said predetermined quantization output levels of said entropy decoders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy coders are different from each other.

26. A voice coding/decoding system as claimed in claim 19, wherein said predetermined numbers of quantizer output levels of said plurality of coders are different from each other, said predetermined quantization characteristics of said plurality of coders are different from each other, the numbers of said predetermined quantization output levels of said entropy coders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy coders are different from each other.

27. A voice coding/decoding system as claimed in claim 26, wherein said predetermined numbers of quantizer output levels of said plurality of decoders are different from each other, said predetermined quantization characteristics of said plurality of decoders are different from each other, the numbers of said predetermined quantization output levels of said entropy decoders in each of said groups are different from each other, and the quantization characteristics in each of said groups of entropy coders are different from each other.

28. A voice coding/decoding system as claimed in claim 19, wherein at least one of said decoders is directly connected to said switching means without connecting through said entropy decoders.

29. A voice coding/decoding system as claimed in claim 19, wherein said plurality of decoders are ADPCM decoders.

30. A voice coding/decoding system as claimed in claim 19, wherein each of said plurality of decoders includes a quantizer.

31. A voice coding/decoding system as claimed in claim 19, wherein said entropy coders are Huffman decoders.

32. A voice coding/decoding system as claimed in claim 19, wherein said entropy decoders are arithmetic decoders.

33. A voice coding/decoding system as claimed in claim 19, wherein said entropy decoders are runlength decoders.

34. A voice coding/decoding system as claimed in claim 19, wherein said entropy decoders are Ziv-lempel decoders.

35. A voice coding/decoding system having a transmitting part for transmitting a coded signal of an input voice signal at a bit rate lower than a predetermined transmission bit rate and a receiving part receiving and decoding the coded signal transmitted from said transmission part, said transmitting part comprising:

a plurality of coders, each having an output, for receiving an input voice signal, and for coding the input voice signal, said coders having predetermined numbers of quantizer output levels and having a predetermined quantization characteristic;

a plurality of groups of entropy coders, each of said entropy coders having an input operatively connected to a corresponding output of said plurality of coders and outputs, said entropy coders having predetermined numbers of quantizer output levels and predetermined probability distributions;

evaluation means, operatively connected to the outputs of said plurality of coders and the outputs of said plurality of entropy coders, for evaluating characteristics of the outputs of said coders and said entropy coders in order to select, from said entropy coders, selected entropy coders having output bit rates lower than the transmission bit rate and to select, from the coders connected to the selected entropy coders, a selected coder having the best output characteristic, by outputting a selecting signal indicating a combination of the selected coder and an entropy coder in the selected entropy coders; and

selecting means, operative connected to the outputs of said plurality of entropy coders and to the output of said evaluation means, for selecting, in response to the selecting signal, a codeword passed through said combination of the selected coder and entropy coder to be transmitted.

36. A voice coding/decoding system as claimed in claim 35, wherein said predetermined numbers of quantizer output levels of said plurality of coders are the same as each other, said predetermined quantization characteristics of said plurality of coders are different from each other, the numbers of said predetermined quantization output levels of said entropy coders are the same as each other, and the quantization characteristics of said entropy coders are different from each other.

37. A voice coding/decoding system as claimed in claim 35, wherein said selecting means comprises multiplexing means for multiplexing the codeword passed through said combination with the selecting signal.

38. A voice coding/decoding system as claimed in claim 35, further comprising at least one coder directly connected to said selecting means without connecting through any entropy coder.

39. A voice coding/decoding system as claimed in claim 35, wherein said plurality of coders comprise adaptive differential pulse code modulation coders.

40. A voice coding/decoding system as claimed in claim 35, wherein each of said plurality of coders includes a quantizer.

41. A voice coding/decoding system as claimed in claim 35, wherein each of said plurality of coders comprises a local decoder for generating a local decoded signal by locally decoding the codeword output from the coder, said local decoded signal being supplied to said evaluation means for extracting a coder having said best characteristic.

42. A voice coding/decoding system as claimed in claim 41, wherein the characteristics of said coders are the summation of the absolute values of the error signals in said local decoded signal.

43. A voice coding/decoding system as claimed in claim 41, wherein the characteristics of said coders are the peak value of the absolute values of the error signals in said local decoded signal.

44. A voice coding/decoding system as claimed in claim 41, wherein the characteristics of said coders are the cepstrum distances in said local decoded signal.

45. A voice coding/decoding system as claimed in claim 35, wherein the characteristics of said coders are signal to noise ratios in said local decoded signal.

46. A voice coding/decoding system as claimed in claim 35, wherein said entropy coders are Huffman coders.

47. A voice coding/decoding system as claimed in claim 35, wherein said entropy coders are arithmetic coders.

48. A voice coding/decoding system as claimed in claim 35, wherein said entropy coders are runlength coders.

49. A voice coding/decoding system as claimed in claim 35, wherein said entropy coders are Ziv-lempel coders.

50. A voice coding/decoding system as claimed in claim 35, wherein said receiving part comprises:

demultiplexing means, operatively connected to the output of said transmitting part through a transmission line, for demultiplexing a received signal from said transmitting part into the codeword and the selecting signal;

switching means, operatively connected to said demultiplexing means, having a single input end for receiving the code word and a plurality of output ends, for passing said codeword from said demultiplexing means to one of said plurality of output ends in response to the selecting signal from said demultiplexing means;

a plurality of entropy decoders, each of said entropy decoders having inputs operatively connected to said plurality of output ends of said switching means and outputs, said entropy decoders having predetermined numbers of quantizer output levels and predetermined probability distributions corresponding to a corresponding entropy coder in said transmitting part; and

a plurality of decoders each having an input, the outputs of said entropy decoders being operatively connected to the inputs of said plurality of decoders, respectively, for decoding the entropy decoded codeword from said entropy decoders, said decoders having predetermined numbers of quantizer output levels and having the predetermined quantization characteristics;

whereby, in accordance with the selecting signal indicating the best combination of a coder and an entropy coder, by the corresponding combination of one of said entropy decoders and one of said coders, the transmitted codeword is decoded.

51. A voice coding/decoding system as claimed in claim 50, wherein said predetermined numbers of quantizer output levels of said plurality of coders are the same as each other, said predetermined quantization characteristics of said plurality of coders are different from each other, the numbers of said predetermined quantization output levels of said entropy coders are the same as each other, and the quantization characteristics of said entropy coders are different from each other.

52. A voice coding/decoding system as claimed in claim 51, wherein said predetermined numbers of quantizer output levels of said plurality of decoders are the same as each other, said predetermined quantization characteristics of said plurality of decoders are different from each other, the numbers of said predetermined quantization output levels of said entropy decoders are the same as each other, and the quantization characteristics of said entropy decoders are different from each other.

53. A voice coding/decoding system as claimed in claim 50, further comprising at least one decoder directly connected to said switching means without connecting through any entropy decoder.

54. A voice coding/decoding system as claimed in claim 50, wherein said plurality of decoders are ADPCM decoders.

55. A voice coding/decoding system as claimed in claim 50, wherein said entropy coders are Huffman decoders.

56. A voice coding/decoding system as claimed in claim 50, wherein said entropy decoders are arithmetic decoders.

57. A voice coding/decoding system as claimed in claim 50, wherein said entropy decoders are runlength decoders.

58. A voice coding/decoding system as claimed in claim 50, wherein said entropy decoders are Ziv-lempel decoders.

59. A voice coding/decoding system having a transmitting part for transmitting a coded signal of an input voice signal at a bit rate lower than a predetermined transmission bit rate and a receiving part for receiving and decoding the coded signal transmitted from said transmission part, said receiving part comprising:

demultiplexing means for receiving and demultiplexing a received signal into a codeword and a selecting signal;

switching means, operatively connected to said demultiplexing means, having an input end for receiving the codeword and having a plurality of output ends, to pass said codeword to a selected one of said plurality of output ends in response to the selecting signal;

a plurality of entropy decoders, each of said entropy decoders having inputs operatively connected to said plurality of output ends of said switching means and outputs, each of said entropy decoders having predetermined numbers of quantizer output levels and predetermined probability distributions corresponding to entropy coders in the transmission part; and

a plurality of decoders each having an input, the outputs of said entropy decoders being operatively connected to the input of one of said plurality of decoders, for decoding an entropy decoded codeword, said decoders having predetermined numbers of quantizer output levels and having predetermined quantization characteristics similar to the transmission part;

whereby, in accordance with the selecting signal directing the switching means to pass the codeword to the selected one of said plurality of output ends, the transmitted codeword is decoded.

BACKGROUND OF THE INVENTION

(1) Field of the Invention

The present invention relates to a voice coding/decoding system having a transmitting part for transmitting a coded signal of an input voice signal and a receiving part for receiving and decoding the coded signal transmitted from the transmission part, and more particularly, to a voice coding/decoding system that does not utilize a buffer memory.

In general, the transmission part has a coder for adaptive differential PCM (ADPCM), adaptive transform coding (ATC) and so forth, and an entropy coder. The coder provides an encoded word which is obtained by compacting the amount of input information. The coded word is then changed to an entropy-coded word by the entropy coder. The entropy-coded word is transmitted from the transmission part to the receiving part. The receiving part conversely decodes the entropy-coded word to reproduce the voice signal.

Accompanied by the developments of digital lines in recent years, a system for compacting the amount of information to obtain coded signals at a high efficiency has been desired in order to effectively utilize the lines.

Also, to store voice information in a voice storage or voice response system, it is important to decrease the capacity of the storing memory. The system for compacting the amount of information and for coding at a high efficiency will satisfy the above requirements.

(2) Description of the Related Art

As a voice signal coding apparatus, a coding transmission apparatus is conventionally known in which a coder and an entropy coder are combined to code voice signals with high efficiency. The coder carries out compacting and coding of voice signals by means of adaptive differential PCM (ADPCM), adaptive transform coding (ATC) and so forth. The entropy coder carries out removing of statistical redundancy in quantized values.

In the coding transmission apparatus for carrying out coding of voice signals at high efficiency, since the voice signals have nonuniform probability of symbols, the statistical characteristics of the output of the coder are changed so that the code lengths, i.e., bit rates, of the entropy-coded results are not constant, and the bit rate may happen to be larger than the allowable transmission bit rate so that the transmission becomes impossible.

Conventionally, to solve the above problem, the entropy encoded result is stored in a buffer memory and is transmitted from the buffer memory through the transmission line. To this end, a buffer control system has been employed in which the entropy-coded results are controlled by changing the quantization characteristics of a quantizer in the coder. This conventional system is described in "On the Information Rate Control in Entropy-coded Speech Transmission Systems", written by M. Copperi, CSELT Rapportitecnici Vol. X-No. 6-DECEMBER, 1982 PP 435-449.

In the above described conventional art, a buffer memory is used to control the entropy coded results. In practice, however, an overflow or an underflow of the buffer memory occurs depending on the characteristics of the coder and the entropy coder.

Further, in the control method by means of the buffer memory, the control is carried out in such a way that the input/output characteristics of the quantizer in the coder is changed so that the statistical characteristics of the quantized output codewords are controlled so as to make the entropies small. Namely, the entropies concentrate to certain values. Therefore, the characteristics of the quantizer itself do not always become optimum, and the characteristic of the coder is not always good.

Further, when there is a transmission error, the error is transmitted in the entropy decoder. Therefore, once a transmission error is generated, the error is superimposed on the next decoded result causing a problem such that the one transmission error influences several samples.

SUMMARY OF THE INVENTION

In view of the above problems in the conventional art, an object of the present invention is to provide a voice coding/decoding system capable of coding and transmitting input voice signal in an optimum state without passing through a buffer memory and without having a negative influence on the coder.

To attain the above object, there is provided, according to the present invention, a voice coding/decoding system having a transmitting part for transmitting a coded signal of an input voice signal at a bit rate lower than a predetermined transmission bit rate and a receiving part for receiving and decoding the coded signal transmitted from the transmission part. The transmitting part comprises a plurality of coders for coding the input voice signals. The coders have different numbers of quantizer output levels, different quantization characteristics, or different numbers of quantizer output levels and different quantization characteristics. The transmitting part also comprises a plurality of groups of entropy coders. The inputs of the entropy coders in each of the groups are connected to the output of one of the plurality of coders. The entropy coders in each of the groups have the same number of quantizer output levels as the number of quantizer output levels in one of the plurality of coders connected to the group, and different probability distributions. The transmitting part further comprises an evaluation part, operatively connected to the local decoder outputs of the plurality of coders and the outputs of the plurality of entropy coders, for evaluating the characteristics of the local decoder outputs of the coders and the entropy coders to extract the entropy coders having output bit rates lower than the transmission bit rate and to extract, from the coders connected to the extracted entropy coders, a coder having the best output characteristic, to output a selecting signal indicating the combination of the selected order and an entropy coder among the extracted entropy coders. The transmitting part still further comprises a selecting part, operatively connected to the outputs of the plurality of entropy coders and to the output of the evaluation part, for selecting, in response to the selecting signal, the codeword passed through the combination of the coder and the entropy coder for transmission.

The receiving part comprises a plurality of decoders for decoding the received codeword, having different numbers of quantizer output levels from each other, different quantization characteristics, or different numbers of quantizer output levels from each other, different quantization characteristics; and a plurality of groups of entropy decoders, the outputs of the entropy decoders in each of the groups being connected to the input of one of the plurality of decoders, the entropy decoders in each of the groups having the same number of quantizer output levels as the number of quantizer output levels in the one of the plurality of coders connected to the group and having different probability distributions; whereby, in accordance with the selecting signal indicating the combination of the selected coder and the selected entropy coder, the transmitted codeword is decoded by the combination of an entropy decoder and a decoder corresponding to the combination of the selected order and the selected entropy order.

BRIEF DESCRIPTION OF THE DRAWINGS

The above object and features of the present invention will be more apparent from the following description of the preferred embodiment with reference to the drawings, wherein:

FIG. 1 is a diagram showing a conventional voice coding and transmission system;

FIG. 2 is a diagram showing the principal construction of a voice coding/decoding system according to an embodiment of the present invention;

FIG. 3 is a diagram showing in detail the transmitting part of the voice coding/decoding system shown in FIG. 2;

FIG. 4 is a diagram showing in detail the receiving part of the coding/decoding system shown in FIG. 2;

FIG. 5 is a flowchart for explaining the operation of the system shown in FIG. 2; and

FIG. 6 is a diagram showing a transmitting part of a voice coding/decoding system according to another embodiment of the present invention; and

FIG. 7 is a diagram showing a receiving part of a voice coding/decoding system according to another embodiment of the present invention.

DESCRIPTION OF THE PREFERRED EMBODIMENT

For better understanding of the present invention, a conventional voice coding transmission system is first described with reference to FIG. 1. In FIG. 1, 71 represents an ADPCM coder, 72 represents an entropy coder, and 73 represents a buffer memory.

As described before, since the voice signals have nonuniform probability of symbols, the statistical characteristics of the output of the coder are changed so that the code lengths, i.e., bit rates, of the entropy-coded results are not constant, and the bit rate may be larger than the transmission bit rate so that the transmission becomes impossible. To solve this problem, the entropy-coded result is stored in the buffer memory 73 and is transmitted from the buffer memory 73 to the transmission line. To this end, a buffer control system has been employed in which the entropy-coded results are controlled by changing the quantization characteristics of a quantizer in the coder. This conventional system is described in "On the Information Rate Control in Entropy-coded Transmission Systems", written by M. Copperi, CSELT Rapportitecnici Vol. X-No. 6-DECEMBER, 1982 PP 435-449.

In the above described conventional art, a buffer memory 73 is used to control the entropy-coded results. In practice, however, an overflow or an underflow of the buffer memory 73 occurs depending on the characteristics of the coder and the entropy coder.

Further, in the control method using part of the buffer memory, the control is carried out in such a way that the input/output characteristics of the quantizer in the coder is changed so that the statistical characteristics of the quantized output codewords are controlled in such a way that the entropy becomes small, namely, the entropy becomes a certain small value. Therefore, the characteristics of the quantizer itself do not always become optimum, and the characteristic of the coder is not always good.

Further, when there is a transmission error, the error is transmitted in the entropy decoder. Therefore, once a transmission error is generated, the error is superimposed on the next decoded result so that there is a problem in that the one transmission error influences several samples.

The above-described problems in the conventional art are solved by the present invention.

Embodiments of the present invention will be described in the following.

FIG. 2 is a diagram showing the principal construction of a voice coding/decoding system according to an embodiment of the present invention. In the figure, the system consists of a transmission part 1 and a receiving part 2. The transmission part 1 includes a plurality of coders Al to An, a plurality of groups GAl to GAn of entropy coder All-Alm, . . . , Anl-Anm, evaluation part 3, and a selecting and multiplexing part 4. Each of the groups GAl to GAn of the entropy coders corresponds to one of the coders Al to An. The evaluation part 3 selects outputs of the entropy coders having bit rates below the transmission bit rate, evaluates the qualities of the coders corresponding to the selected entropy coders, selects an optimum combination of a coder and an entropy coder, and generates a selecting signal indicating the combination of the selected coder and the selected entropy coder. The selecting and multiplexing part 4 multiplexes the selected codeword and the selecting signal.

The plurality of coders Al to An have different numbers of quantizer output levels (bit lengths). Alternatively, the plurality of coders Al to An may have different quantization characteristics. As an another alternative, the plurality of coders Al to An may have different numbers of quantizer output levels and different quantization characteristics. As an another alternative, the plurality of coders Al to An may have different numbers of quantizer output levels and different quantization characteristics. Each of the coders corresponds to one of the groups GAl to GAn. In each of the groups GAl to GAn, a plurality of entropy coders have different probability distributions. The plurality of entropy coders are operated in parallel. The evaluation part 3 selects the outputs of the entropy coders having bit rates lower than the transmission bit rate from all of the outputs of the entropy coders All to Anm for each frame of input voice signals, evaluates the qualities of the local decoded signals from all of the coders corresponding to the entropy coders from which the selected outputs are generated, and selects a codeword having the optimum quality. The evaluation part 3 also generates a selecting signal indicating the combination of the coder and the entropy coder from which the codeword is selected. The selecting signal is applied to the selecting and multiplexing part 4. The selecting and multiplexing part 4 multiplexes the selected codeword and the selecting signal to transmit through the transmission line to the receiving part 2.

The receiving part 2 includes a demultiplexing part 5, a switching part 7, a plurality of groups GBl to GBn of a plurality of entropy decoders Bll-Blm, . . . , Bn1-Bnm, and a plurality of decoders Bl to Bn each corresponding to one of the plurality of groups Bll-Blm, . . . , Bnl-Bnm. The demultiplexing part 5 demultiplexes the multiplexed signal received from the transmitting part 1 through the transmission line into the codeword and the selecting signal. The switching part 7 passes, in response to the demultiplexed selecting signal from the demultiplexing part 5, the demultiplexed codeword to the combination of the entropy decoder and the decoder corresponding to the combination of the coder and the entropy coder selected in the transmitting part 1. The thus selected entropy decoder carries out an entropy decoding in accordance with the probability distribution of the entropy decoder, and the selected decoder decodes, in accordance with the number of the quantizer output levels, the quantization characteristics, or the number of the quantizer output levels and the quantization characteristics, the output of the selected entropy decoder to reproduce the voice signal.

Since the voice signal has nonuniform probability of symbols, the statistical characteristics of the codewords output from a coder are different for respective frames. Therefore, the codeword lengths of the entropy coded results are different for respective frames depending on the number of the quantizer output levels, the quantization characteristics, or the number of the quantizer output levels and the quantization characteristics. In view of these facts, according to the present invention, a plurality of coders are provided and each of the coders is made to correspond to one of the groups of a plurality of entropy coders having different probability distributions. And the optimum combination of a coder and an entropy coder is selected to transmit a codeword.

In the transmitting part 1, an input voice signal is input in parallel to the coders Al to An having different number of quantizer output levels, quantization characteristics, or number of quantizer output levels and quantization characteristics. Then coders Al to An carry out the coding simultaneously to output codewords and local decoded signals having different numbers of quantizer output levels, different quantization characteristics, or different numbers of quantizer output levels and different quantization characteristics. The codewords have statistical nonuniformity so that there are redundancies due to the statistical nonuniformity. To remove the redundancies, the plurality of entropy decoders All to Anm respectively having the different probability distributions are employed to carry out entropy codings.

In the evaluation part 3, the codeword lengths of the entropy-coded results and the qualities of the local decoded signals are evaluated for each frame so that the codewords, having codeword lengths