Low-bit-rate speech coder using LPC data reduction processing

• US 4,975,956 A
• Filed: 07/26/1989
• Issued: 12/04/1990
• Est. Priority Date: 07/26/1989
• Status: Expired due to Fees

• Current Assignee: ITT Corporation (ITT, Inc.)
• Original Assignee: ITT Corporation (ITT, Inc.)
• Inventors: Liu, Yu J., Rothweiler, Joseph
• Primary Examiner(s): Kemeny, Emanuel S.
• Application Number:
  US07/385,644
• Time in Patent Office: 496 Days
• Field of Search: 381/29-53, 375/122
• US Class Current: 704/219
• CPC Class Codes: G10L 15/07 to the speaker
• CPC Class Codes: G10L 19/038 Vector quantisation, e.g. TwinVQ audio
• CPC Class Codes: G10L 19/06 Determination or coding of the spectral characteristics, e.g. of the short-term prediction coefficients
• CPC Class Codes: G10L 19/07 Line spectrum pair [LSP] vocoders

Abstract

A speech coder employs vector quantization of LPC parameters, interpolation, and trellis coding for improved speech coding at low bit rates (400 bps). The speech coder has an LPC analysis module for converting input speech to LPC parameters, an LSP conversion module for converting LPC parameters into line spectrum frequencies (LSP) data, and a vector quantization and interpolation (VQ/I) module for encoding the LSP data into vector indexes for transmission by applying LPC spectral amplitude as weighting coefficients to the LSP data. The VQ/I module outputs one vector index for every two LPC frames in order to reduce the transmission bit rate, and the omitted frames are interpolated on the receiving end. A decoder correspondingly decodes incoming indexes to LPC parameters and synthesizes them into output speech. Trellis coders with an adaptive tracking function encode the pitch and gain parameters of the LPC frames. A universal codebook stores codewords according to a plurality of accents. The speech coder automatically identifies a speaker'"'"'s accent and selects the corresponding vocabulary of codewords in order to more intelligibly encode and decode the speaker'"'"'s speech.

Claims

1. A speech coder system comprising an encoder for encoding input speech signals for transmission over a communication channel at a low bit rate, said encoder including:

an LPC analysis module for converting input speech to linear predictive coding (LPC) parameters for each of a sequence of LPC frames;

an LSP conversion module for converting the LPC parameters to corresponding line spectrum frequencies (LSP) data, and

a vector quantization and interpolation (VQ/I) encoder module for encoding the LSP data into vector indexes for transmission,

wherein said VQ/I encoder module includes vector quantizing means for computing weighting coefficients for the LSP data, each of which is the product of LPC spectral amplitude and a predetermined factor, and means for applying said weighting coefficients to said LSP data.

2. A speech coder system according to claim 1, wherein said VQ/I encoder module outputs one vector index for every two LPC frames in order to provide a low transmission bit rate.

5. A speech coder system according to claim 1, further comprising a decoder for decoding incoming low-bit-rate signals transmitted by said encoder over a communication channel, said decoder including:

a vector quantization and interpolation (VQ/I) decoder module for decoding said incoming signals into corresponding line spectrum frequencies (LSP) data;

an LSP conversion module for converting the LSP data to corresponding LPC parameters; and

an LPC synthesis module for converting the LPC parameters into output speech.

6. A speech coder system according to claim 5, wherein said VQ/I encoder module outputs one vector index for every two LPC frames in order to provide a low transmission bit rate, and said VQ/I decoder module generates LPC parameters for every two LPC frames by interpolating an LPC frame from the vector indexes for adjacent frames.

7. A speech coder system according to claim 5, wherein said LPC parameters include pitch and gain data, and said encoder further comprises trellis coders for respectively encoding said pitch and gain data into pitch and gain bits for transmission, and said decoder further comprises pitch and gain decoders for respectively restoring said pitch and gain data from said pitch and gain bits.

8. A speech coder system according to claim 7, wherein said trellis coders of said encoder provide for adaptive tracking of respective sequences of pitch and gain data, and for selecting pitch and gain bits corresponding to minimum distortion in the respective sequences, and said trellis decoders of said decoder respectively restore said pitch and gain data from said selected pitch and gain bits.

9. A speech coder system according to claim 5, wherein said encoder further comprises a codebook for storing a vocabulary of codewords in the form of respectively defined vector indexes corresponding to sound to be spoken as input speech signals to said encoder, and means for comparing the quantized vector indexes output from said VQ/I encoder module to the codewords of said codebook and outputting the defined vector indexes corresponding thereto with a minimum distortion for transmission.

10. A speech coder system according to claim 9, wherein said codebook is a universal codebook for storing a plurality of vocabulary modules each corresponding to a vocabulary of codewords in the form of defined standard vector indexes corresponding to wound spoken as input speech signals in a respective spoken accent, and wherein said encoder includes speech identifying means for receiving a sample of input speech signals from a speaker and determining a corresponding accent of the speaker, selection means for selecting a vocabulary module of said universal codebook corresponding to said determined accent, and comparing means for comparing the quantized vector indexes output from said VQ/I encoder module in response to input speech in said accent to said selected vocabulary module and for outputting the defined standard vector indexes corresponding thereto.

11. A speech coder system according to claim 10, wherein said decoder includes a corresponding universal codebook for storing a plurality of vocabulary modules each having a vocabulary of codewords in the form of defined accented vector indexes corresponding to a respective spoken accent, wherein said speech identifying means of said encoder provides a vocabulary code signal identifying the accent of the speaker for transmission, and wherein said decoder includes selection means for receiving the transmitted vocabulary code signal and selecting a vocabulary module of said universal codebook corresponding to the speaker'"'"'s accent, and comparing means for comparing the transmitted standard vector indexes to said selected vocabulary module and for outputting corresponding accented vector indexes to said VQ/I decoder module for decoding into output speech in said accent.

12. A speech coder system according to claim 1, wherein said LSP conversion module includes a sum filter and a difference filter for converting the LPC parameters to a sum polynomial and a difference polynomial, respectively, having real roots only, and means for solving the roots of the sum and difference polynomials.

13. A speech coder system according to claim 12, wherein said LPC analysis module outputs LPC parameters having N LPC coefficients, and said means for solving of said LSP conversion module includes means for computing respective Nth order polynomials from the LPC coefficients, reducing the respective polynomial orders to N/2, solving the respective polynomial roots, and combining and ordering all N roots.

14. A speech coder system according to claim 5, wherein said LPC analysis module of said encoder outputs LPC parameters having N LPC coefficients, and said LSP conversion module of said decoder includes means for obtaining LPC coefficients from said LSP coefficients.

3. A speech coder system according to claim wherein said LPC parameters include pitch and gain data, and said encoder further comprises trellis coders for respectively encoding said pitch and gain data into pitch and gain bits for transmission.

4. A speech coder system according to claim 3, wherein said trellis coders provide for adaptive tracking of respective sequences of pitch and gain data, and for selecting pitch and gain bits corresponding to minimum distortion in the respective sequences.