WO1993006590A1 - A speech coding device - Google Patents

A speech coding device Download PDF

Info

Publication number
WO1993006590A1
WO1993006590A1 PCT/BE1991/000067 BE9100067W WO9306590A1 WO 1993006590 A1 WO1993006590 A1 WO 1993006590A1 BE 9100067 W BE9100067 W BE 9100067W WO 9306590 A1 WO9306590 A1 WO 9306590A1
Authority
WO
WIPO (PCT)
Prior art keywords
code
input
word
code word
words
Prior art date
Application number
PCT/BE1991/000067
Other languages
French (fr)
Inventor
Gao Yang
Original Assignee
Lernout & Hauspie Speechproducts
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Lernout & Hauspie Speechproducts filed Critical Lernout & Hauspie Speechproducts
Priority to EP91915344A priority Critical patent/EP0558492A1/en
Priority to JP3514292A priority patent/JPH06502928A/en
Priority to PCT/BE1991/000067 priority patent/WO1993006590A1/en
Publication of WO1993006590A1 publication Critical patent/WO1993006590A1/en

Links

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
    • G10L19/08—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
    • G10L19/08—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters
    • G10L19/12—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters the excitation function being a code excitation, e.g. in code excited linear prediction [CELP] vocoders
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L2019/0001—Codebooks
    • G10L2019/0011—Long term prediction filters, i.e. pitch estimation
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L2019/0001—Codebooks
    • G10L2019/0013—Codebook search algorithms
    • G10L2019/0014—Selection criteria for distances

Definitions

  • the invention relates to a speech coding device having a first input for receiving a first error word determined upon an ideal and an estimated excitation word themselves determined from a speech ' signal, said first input being connected to a second input of a first unit, a third input of which being connected to a code word generator, said first unit being provided for determining a second error word upon inputted words, an output of said first unit being connected to a filter unit which is close looped with said code word generator, said code word generator being provided for generating a series of code words, said coding device comprises processing means provided for determining upon words output by said filter unit an energy value and for selecting among said code words the one having produced the lowest energy value.
  • Such a speech coding device is known from the article of H. Hassanein, A Brind' Amour and K. Bryden entitled "A 4800 bps
  • the first error word input into the speech coding device originates for example from a weighting filter unit.
  • the code word generator comprises a code book having a first respectively a second part wherein periodic code words for voiced speech signals respectively non-periodic code words for unvoiced speech signals are stored. The choice between a first or a second code word being realized upon an analysis of the input speech signal in order to determine whether the input speech signal is voice or unvoiced. Such an analysis is done by using the
  • LTP Long Term Prediction
  • word be it an error word or a code word means a sequence of binary samples.
  • a drawback of the known speech coding device is that the generated code word is fully dependent of the choice made between a voiced or unvoiced speech signal. Even if the choice is correctly made, it does not necessarily imply that the best code word is found in the selected code book. That best code word could as well be found in the non-chosen code book. When the choice voiced- unvoiced is not correctly done, this could also lead to a wrong choice of code book.
  • the basic problem in chosing between a voiced and unvoiced speech signal is that a lot of speech signals are in fact most of the time built up by a mixture of voiced and unvoied signals.
  • a speech coding device is therefore characterized in that said code word generator being provided for generating a first series of first code words having a substantially periodic character with a period associated with an input pitch period , said code word generator being also provided for genera ⁇ ting a second series of second code word having a substantially non- periodic character. Since the code word generator generates now as well first as second code words it is no longer necessary to make a choice between voiced and unvoiced speech so that errors caused by an erroneous choice are avoided.
  • a first preferred embodiment of a speech coding device is characterized in that said code word generator is provided for generating a third series of third code words having a substantially periodic character with a period correspon- ding to a fraction m (m £ R) or a multiple r (r ⁇ . R ) of said input pitch value.
  • the pitch value is generally determined by an LTP analysis.
  • Figure 1 shows an embodiment of a coding device according to the invention.
  • Figure 2 respectively 3 shows a first respectively a second series of code words.
  • the illustrated speech coding device comprises a first input 1 for receiving a first error word.
  • This first error word is for example determined by computing the difference between an ideal and an estimated excitation word obtained after processing a speech signal which originates for example from a human voice.
  • the first input 1 is connected to a second input of a first unit 2.
  • the first unit is for example formed by a subtracting unit.
  • An output of said first unit 2 is connected to a filter unit 3, which is close looped via a unit provided for determining the energy value of the output signal of said filter unit 3, and a minimization unit 5 with a code word generator 6.
  • An output of the code word generator 6 is connected via a gain multiplying element 7 with a third input of the first unit 2.
  • An input first error word is supplied to the first unit 2 which also receives an estimated word. Said estimated word being formed by multiplying a code word output by the code word generator 6 with a gain value. The latter operation being realized by element 7.
  • the first unit deducts the estimated word from the first word and thus forms a second error word which is supplied to the filter unit 3 in order to form a third error word.
  • the unit compu ⁇ tes the energy value of input third error word. This energy value is then input into the minimization unit 5 which controls the code word generator 6.
  • the code word generator 6 is provided for generating a first series of first code words as well as a second series of second code words. For the sake of clarity, the first code word will first be considered.
  • Figure 2 shows an example of a first series of first code words.
  • L 50 samples
  • P 20 samples.
  • the distance between successive * non-zero bits in the first code words is each time 20 samples, i.e. the pitch length P.
  • Each subsequent (b, c,..i) first code word of said first series is then each time obtained by time directional shifting the non-zero bits over one sample period.
  • the first code word illustrated under b in figure 2 has non-zero bits at sample position 1, 21 and 41 while the last first code word (i in figure 2) has only one non-zero bit at sample position 49.
  • P * L however if P> L there is of course only one non-zero bit in each first code word.
  • each first code word of said first series is successively supplied to said third input of the first unit in order to determine the second and third error word.
  • the code word generator is further provided for selecting among the first code words, the one having produced the third error word with the lowest energy value. This is for example realized each time when comparing the energy value obtained by the considered code word with a temporarily memorized lowest energy obtained by a preceding first code word during a same operation and by overruling the memorized lowest energy value if the latter is larger than the one obtained by the considered first code word.
  • the code word generator is also provided for generating a second series of second code words upon a received predetermined code value Q indicating the number of bits quantified as non-zero bits that have to be considered in said second code words.
  • the signal e_ illustrates an example of an analogous form of a first error word
  • the distance between the successive non-zero bits being determined by L/Q, where L is again the subframe length.
  • the code word generator comprises processing means for determining upon said second code word f_ and said first error word a first resp. a second candidate word f_ resp. f . of which an example is shown in figure 3. That first candidate word f, is for example obtained by multiplying f_ with a binary form of e_, while
  • the second candidate word f is obtained by inversing the first non-zero bit value of the first candidate word.
  • the first and the second candidate words are now successively input into the first device and the filter unit in order to obtain each time a third error word e, . and e, _.
  • the processing 1-5 means of the code word generator being also provided for comparing e. . with e, _ and for selecting among both third error words, the one having the lowest energy value.
  • the candidate word having produced the third error word with the lowest energy value is then considered for further processing.
  • the first candidate word is chosen for further processing because its related lower energy value indicates that this candidate word is closer to the first error word.
  • a third candidate word f now generated by taking
  • the energy value e, . of the remaining candidate word is compared
  • the code word generator of the speech coding device is preferably provided for generating a third series of third code words.
  • the third code words are built up in an analogous manner as the first code words but the distance between successive bits quantified as non-zero bits now equals a fraction m (m lR) of a multiple r .r _£ ⁇ R) of the input pitch value P.
  • m 2 and thus the distance between successive non-zero bits is P/2.
  • the third code words are considered in an analogous manner as the one of the first series.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Transmission Systems Not Characterized By The Medium Used For Transmission (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Abstract

A speech coding device having a first input for receiving a first error word determined upon an ideal and an estimated excitation word itself determined from a speech signal, said first input being connected to a second input of a first unit, a third input of which being connected to a code word generator, an output of said first unit being connected to a filter unit which is close looped with said code word generator, said said code word generator being provided for generating a first series of first code words having a substantially periodic character with a period associated with an input pitch period, said code word generator being also provided for generating a second series of second code words having a substantially non-periodic character, said coding device comprises processing means provided for determining upon words output by said filter unit an energy value and for selecting among said code words the one having produced the lowest energy value.

Description

"A speech coding device"
The invention relates to a speech coding device having a first input for receiving a first error word determined upon an ideal and an estimated excitation word themselves determined from a speech ' signal, said first input being connected to a second input of a first unit, a third input of which being connected to a code word generator, said first unit being provided for determining a second error word upon inputted words, an output of said first unit being connected to a filter unit which is close looped with said code word generator, said code word generator being provided for generating a series of code words, said coding device comprises processing means provided for determining upon words output by said filter unit an energy value and for selecting among said code words the one having produced the lowest energy value.
Such a speech coding device is known from the article of H. Hassanein, A Brind' Amour and K. Bryden entitled "A 4800 bps
CELP Vocoder with an Improved Excitation" and published in the papers of the International Mobile Satellite conference of Ottowa, 1990.
The first error word input into the speech coding device originates for example from a weighting filter unit. The code word generator comprises a code book having a first respectively a second part wherein periodic code words for voiced speech signals respectively non-periodic code words for unvoiced speech signals are stored. The choice between a first or a second code word being realized upon an analysis of the input speech signal in order to determine whether the input speech signal is voice or unvoiced. Such an analysis is done by using the
LTP (Long Term Prediction) gain and a predetermined threshold value. The term "word" be it an error word or a code word means a sequence of binary samples. A drawback of the known speech coding device is that the generated code word is fully dependent of the choice made between a voiced or unvoiced speech signal. Even if the choice is correctly made, it does not necessarily imply that the best code word is found in the selected code book. That best code word could as well be found in the non-chosen code book. When the choice voiced- unvoiced is not correctly done, this could also lead to a wrong choice of code book. The basic problem in chosing between a voiced and unvoiced speech signal, is that a lot of speech signals are in fact most of the time built up by a mixture of voiced and unvoied signals.
It is an object of the present invention to realize a speech coding device wherein an adequate code word is generated, independently whether the processed signal is voiced or unvoiced, a mixture of voiced-unvoiced, a plosive sound or any other type of speechlike signal.
A speech coding device according to the invention is therefore characterized in that said code word generator being provided for generating a first series of first code words having a substantially periodic character with a period associated with an input pitch period , said code word generator being also provided for genera¬ ting a second series of second code word having a substantially non- periodic character. Since the code word generator generates now as well first as second code words it is no longer necessary to make a choice between voiced and unvoiced speech so that errors caused by an erroneous choice are avoided.
A first preferred embodiment of a speech coding device according to the invention is characterized in that said code word generator is provided for generating a third series of third code words having a substantially periodic character with a period correspon- ding to a fraction m (m £ R) or a multiple r (rξ. R ) of said input pitch value. The pitch value is generally determined by an LTP analysis.
Due to the fact that by this LTP analysis the pitch value is sometime not correctly determined, an error in the pitch value automatically would lead to erroneous first code words. The provision of said third code words enables to take into account errors made in determining the pitch value. The invention will now be described in more details by means of the drawings, in which :
Figure 1 shows an embodiment of a coding device according to the invention. Figure 2 respectively 3 shows a first respectively a second series of code words.
The illustrated speech coding device comprises a first input 1 for receiving a first error word. This first error word is for example determined by computing the difference between an ideal and an estimated excitation word obtained after processing a speech signal which originates for example from a human voice.
The first input 1 is connected to a second input of a first unit 2.
The first unit is for example formed by a subtracting unit. An output of said first unit 2 is connected to a filter unit 3, which is close looped via a unit provided for determining the energy value of the output signal of said filter unit 3, and a minimization unit 5 with a code word generator 6. An output of the code word generator 6 is connected via a gain multiplying element 7 with a third input of the first unit 2. An input first error word is supplied to the first unit 2 which also receives an estimated word. Said estimated word being formed by multiplying a code word output by the code word generator 6 with a gain value. The latter operation being realized by element 7. The first unit deducts the estimated word from the first word and thus forms a second error word which is supplied to the filter unit 3 in order to form a third error word. The unit compu¬ tes the energy value of input third error word. This energy value is then input into the minimization unit 5 which controls the code word generator 6. The code word generator 6 is provided for generating a first series of first code words as well as a second series of second code words. For the sake of clarity, the first code word will first be considered. Figure 2 shows an example of a first series of first code words. Consider for example the length L of the subframe being L = 50 samples, and the pitch length P = 20 samples. The pitch length is for example determined by means of a LTP (Long Term Prediction) analysis. With L = 50 and P 20, the initial first code word (a in figure 2) in the first series has a bit quantified as non-zero bit at sample position 0, 20 and 40. Thus, the distance between successive * non-zero bits in the first code words is each time 20 samples, i.e. the pitch length P. Each subsequent (b, c,..i) first code word of said first series is then each time obtained by time directional shifting the non-zero bits over one sample period. Thus, for example, the first code word illustrated under b in figure 2 has non-zero bits at sample position 1, 21 and 41 while the last first code word (i in figure 2) has only one non-zero bit at sample position 49. In the example illustrated in figure 2, P* L, however if P> L there is of course only one non-zero bit in each first code word.
In operating the speech coding device according to the invention, each first code word of said first series is successively supplied to said third input of the first unit in order to determine the second and third error word. The code word generator is further provided for selecting among the first code words, the one having produced the third error word with the lowest energy value. This is for example realized each time when comparing the energy value obtained by the considered code word with a temporarily memorized lowest energy obtained by a preceding first code word during a same operation and by overruling the memorized lowest energy value if the latter is larger than the one obtained by the considered first code word. By storing that lowest energy value there is also stored the sequence number of the considered first code word or another index identifying that sequence number, in order to identify that first code word having produced a third error word with the lowest energy value. The code word generator is also provided for generating a second series of second code words upon a received predetermined code value Q indicating the number of bits quantified as non-zero bits that have to be considered in said second code words. In figure 3, the signal e_ illustrates an example of an analogous form of a first error word, while f_ illustrates an example of a second code word wherein Q = 6 and d = 1 , d being a shift value indicating the position of a first non-zero bit in the second code word. The distance between the successive non-zero bits being determined by L/Q, where L is again the subframe length.
The code word generator comprises processing means for determining upon said second code word f_ and said first error word a first resp. a second candidate word f_ resp. f . of which an example is shown in figure 3. That first candidate word f, is for example obtained by multiplying f_ with a binary form of e_, while
10 the second candidate word f . is obtained by inversing the first non-zero bit value of the first candidate word.
The first and the second candidate words are now successively input into the first device and the filter unit in order to obtain each time a third error word e, . and e, _. The processing 1-5 means of the code word generator being also provided for comparing e. . with e, _ and for selecting among both third error words, the one having the lowest energy value. The candidate word having produced the third error word with the lowest energy value is then considered for further processing. 20 Suppose now that e, . < e_ _. Under those circumstan¬ ces, the first candidate word is chosen for further processing because its related lower energy value indicates that this candidate word is closer to the first error word.
A third candidate word f ,. now generated by taking
2-5 the first candidate word and inversing the second non-zero bit. A third error word e_ - is now determined by considering the third candida-
3.3 te word. The energy value of e_ .. is compared with e, . and again the lowest one is selected. The candidate word having generated the lowest energy value is then again considered for determining the
30 fourth candidate word and the process described herebefore is repeated until all Q non-zero bits have been considered, i.e. until the Q candi¬ date word has been generated and considered.
After having considered all Q generated candidate words, the energy value e, . of the remaining candidate word is compared
^5 with the energy value of the first code word selected by considering the series of first code words. Among both energy values, there is now chosen the one having the lowest value. The first code word or the candidate word having the lowest energy value is then selected and this word is the one output by the speech coding unit. From the description set out herebefore, it will be clear that, contrary to what is done in the known coding devices, it is not necessary to make a selection among voiced and unvoiced speech in order to determine a first code word or a second code word. In the description given herebefore only one Q value has been considered for determining the candidate words. However it will be clear that several Q values can be considered. For each consideration of Q value the described process is then repeated and of course the candidate words issued for each Q value are then compared with each other on the basis of the lowest energy value and the one with the lowest energy value is chosen.
It could happen that the pitch value determined by the LTP analysis is not the correct one. This would of course have consequences for the first series of first code words. In order to provide a solution for this problem, the code word generator of the speech coding device according to the present invention is preferably provided for generating a third series of third code words. The third code words are built up in an analogous manner as the first code words but the distance between successive bits quantified as non-zero bits now equals a fraction m (m lR) of a multiple r .r _£ϊR) of the input pitch value P. In a preferred embodiment m = 2 and thus the distance between successive non-zero bits is P/2. The third code words are considered in an analogous manner as the one of the first series. However computation time can be saved by not considering the last third code words of the third series. Since the third code words are, just as the first code words, obtained by shifting the non-zero bits, the last third code words equals the first code words so that it would be redundant to consider them again.
Of course a code word generator provided for genera¬ ting beside a first and a third series also a fourth and even further series each time having a different m or r value could also be realized.

Claims

1. A speech coding device having a first input for receiving a first error word determined upon an ideal and an estimated excitation word itself determined from a speech signal, said first input being connected to a second input of a first unit, a third input of which being connected to a code word generator, said first unit being provided for determining a second error word upon inputted words, an output of said first unit being connected to a filter unit which is close looped with said code word generator, said code word generator being provided for generating a series of code words, said coding device comprises processing means provided for determining upon words output by said filter unit an energy value and for selecting among said code words the one having produced the lowest energy value, characterized in that said code word generator being provided for generating a first series of first code words having a substantially periodic character with a period associated with an input pitch period, said code word generator being also provided for generating a second series of second code words having a substantially non-periodic character.
2. A speech coding device as claimed in claim 1, characterized in that said code word generator is provided for generating a third series of third code words having a substantially periodic charac¬ ter with a period corresponding to a fraction m (m fc _R.) of said input pitch value.
3. A speech coding device as claimed in claim 2, characterized in that said code word generator is provided for generating a third series of third code words having a substantially periodic charac¬ ter with a period corresponding to a a multiple r (r £ fO of said input pitch value.
PCT/BE1991/000067 1991-09-20 1991-09-20 A speech coding device WO1993006590A1 (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
EP91915344A EP0558492A1 (en) 1991-09-20 1991-09-20 A speech coding device
JP3514292A JPH06502928A (en) 1991-09-20 1991-09-20 audio coding element
PCT/BE1991/000067 WO1993006590A1 (en) 1991-09-20 1991-09-20 A speech coding device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/BE1991/000067 WO1993006590A1 (en) 1991-09-20 1991-09-20 A speech coding device

Publications (1)

Publication Number Publication Date
WO1993006590A1 true WO1993006590A1 (en) 1993-04-01

Family

ID=3885301

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/BE1991/000067 WO1993006590A1 (en) 1991-09-20 1991-09-20 A speech coding device

Country Status (3)

Country Link
EP (1) EP0558492A1 (en)
JP (1) JPH06502928A (en)
WO (1) WO1993006590A1 (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0137532A2 (en) * 1983-08-26 1985-04-17 Koninklijke Philips Electronics N.V. Multi-pulse excited linear predictive speech coder
EP0279451A2 (en) * 1987-02-20 1988-08-24 Fujitsu Limited Speech coding transmission equipment

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0137532A2 (en) * 1983-08-26 1985-04-17 Koninklijke Philips Electronics N.V. Multi-pulse excited linear predictive speech coder
EP0279451A2 (en) * 1987-02-20 1988-08-24 Fujitsu Limited Speech coding transmission equipment

Also Published As

Publication number Publication date
JPH06502928A (en) 1994-03-31
EP0558492A1 (en) 1993-09-08

Similar Documents

Publication Publication Date Title
US8364473B2 (en) Method and apparatus for receiving an encoded speech signal based on codebooks
AU683126B2 (en) Linear prediction coefficient generation during frame erasure or packet loss
EP0696026B1 (en) Speech coding device
US5138661A (en) Linear predictive codeword excited speech synthesizer
CA1337217C (en) Speech coding
US6249758B1 (en) Apparatus and method for coding speech signals by making use of voice/unvoiced characteristics of the speech signals
MY119252A (en) Depth-first algebraic-codebook search for fast coding of speech
EP0232456A1 (en) Digital speech processor using arbitrary excitation coding
US5488704A (en) Speech codec
EP0654909A1 (en) Code excitation linear prediction encoder and decoder
US7302387B2 (en) Modification of fixed codebook search in G.729 Annex E audio coding
US6629070B1 (en) Voice activity detection using the degree of energy variation among multiple adjacent pairs of subframes
EP0784846B1 (en) A multi-pulse analysis speech processing system and method
US5642368A (en) Error protection for multimode speech coders
US5875423A (en) Method for selecting noise codebook vectors in a variable rate speech coder and decoder
EP0578436B1 (en) Selective application of speech coding techniques
JPH1097294A (en) Voice coding device
WO1993006590A1 (en) A speech coding device
RU2667462C1 (en) Method of recognizing low-speed speech coding protocols
EP0745972B1 (en) Method of and apparatus for coding speech signal
EP0903729B1 (en) Speech coding apparatus and pitch prediction method of input speech signal
RU2610285C1 (en) Method of detecting low-rate encoding protocols
AU682505B2 (en) Vector coding process, especially for voice signals
JP2700974B2 (en) Audio coding method
US6289307B1 (en) Codebook preliminary selection device and method, and storage medium storing codebook preliminary selection program

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AU CA JP US

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): AT BE CH DE DK ES FR GB GR IT LU NL SE

WWE Wipo information: entry into national phase

Ref document number: 1991915344

Country of ref document: EP

WWP Wipo information: published in national office

Ref document number: 1991915344

Country of ref document: EP

NENP Non-entry into the national phase

Ref country code: CA

WWW Wipo information: withdrawn in national office

Ref document number: 1991915344

Country of ref document: EP