Variable rate vocoder

• US 5,414,796 A
• Filed: 01/14/1993
• Issued: 05/09/1995
• Est. Priority Date: 06/11/1991
• Status: Expired due to Term

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Lee, Chong U., Jacobs, Paul E., Gilhousen, Klein S., Tsai, Ming-Chang, Lam, S. Katherine, Gardner, William R.
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): HAFIZ, TARIQ R
• Application Number:
  US08/004,484
• Time in Patent Office: 845 Days
• Field of Search: 381/30-31, 381/35-36, 381/38, 381/40-41, 381/47, 395/2, 395/2.1, 395/2.3, 395/2.31, 395/2.33, 375/27, 375/122
• US Class Current: 704/221
• CPC Class Codes: G10L 19/005 Correction of errors induced by the transmission channel, if related to the coding algorithm
• CPC Class Codes: G10L 19/012 Comfort noise or silence coding
• CPC Class Codes: G10L 19/12 the excitation function being a code excitation, e.g. in code excited linear prediction [CELP] vocoders
• CPC Class Codes: G10L 19/22 Mode decision, i.e. based on audio signal content versus external parameters
• CPC Class Codes: G10L 19/24 Variable rate codecs, e.g. for generating different qualities using a scalable representation such as hierarchical encoding or layered encoding
• CPC Class Codes: G10L 2025/786 Adaptive threshold
• CPC Class Codes: G10L 25/78 Detection of presence or absence of voice signals switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10
• CPC Class Codes: H04B 1/66 for reducing bandwidth of signals; for improving efficiency of transmission H04B1/68 takes precedence
• CPC Class Codes: H04J 3/1688 the demands of the users being taken into account after redundancy removal, e.g. by predictive coding, by variable sampling reducing bandwidth of signals in general H04B1/66; in PCM-systems H04B14/046; removal of redundancy in telegraph communication H03M7/30
• CPC Class Codes: H04L 1/0014 by adapting the source coding
• CPC Class Codes: H04L 1/0017 where the mode-switching is based on Quality of Service requirement
• CPC Class Codes: H04L 1/0041 Arrangements at the transmitter end
• CPC Class Codes: H04L 1/0046 Code rate detection or code type detection H04L1/0038 takes precedence; detection of the data rate H04L25/0262; for packet format H04L1/0091
• CPC Class Codes: H04L 1/0057 Block codes H04L1/0061, H04L1/0064 take precedence

Abstract

An apparatus and method for performing speech signal compression, by variable rate coding of frames of digitized speech samples. The level of speech activity for each frame of digitized speech samples is determined and an output data packet rate is selected from a set of rates based upon the determined level of frame speech activity. A lowest rate of the set of rates corresponds to a detected minimum level of speech activity, such as background noise or pauses in speech, while a highest rate corresponds to a detected maximum level of speech activity, such as active vocalization. Each frame is then coded according to a predetermined coding format for the selected rate wherein each rate has a corresponding number of bits representative of the coded frame. A data packet is provided for each coded frame with each output data packet of a bit rate corresponding to the selected rate.

Claims

1. A method of speech signal compression, by variable rate coding of frames of digitized speech samples, comprising the steps of:

determining a level of speech activity for a frame of digitized speech samples;

selecting an encoding rate from a set of rates based upon said determined level of speech activity for said frame;

coding said frame according to a coding format of a set of coding formats for said selected rate wherein each rate has a corresponding different coding format and wherein each coding format provides for a different plurality of parameter signals representing said digitized speech samples in accordance with a speech model; and

generating for said frame a data packet of said parameter signals.

2. The method of claim 1 wherein said step of determining said level of frame speech activity comprises the steps of:

measuring speech activity in said frame of digitized speech samples;

comparing said measured speech activity with at least one speech activity threshold level of a predetermined set of activity threshold levels; and

adaptively adjusting in response to said comparison at least one of said at least one speech activity threshold levels with respect to a level of activity of a previous frame of digitized speech samples.

3. The method of claim 1 further comprising the steps of:

providing a rate command indicative of a preselected encoding rate for said frame; and

modifying said selected encoding rate to provide said preselected encoding rate for coding of said frame at said preselected encoding rate.

4. The method of claim 3 wherein said preselected rate is less than a predetermined maximum rate, said method further comprising the steps of:

providing an additional data packet; and

combining said data packet with said additional data packet within a transmission frame for transmission.

5. The method of claim 1 wherein said step of providing said data packet of said parameter signals comprises:

generating a variable number of bits to represent linear predictive coefficient (LPC) vector signals of said frame of digitized speech samples, wherein said variable number of bits representing said LPC vector signals is responsive to said measured speech activity level;

generating a variable number of bits to represent pitch vector signals of said frame of digitized speech samples, wherein said variable number of bits representing said pitch vector signals is responsive to said measured speech activity level; and

generating variable number of bits to represent codebook excitation vector signals of said frame of digitized speech samples, wherein said variable number of bits representing said codebook excitation vector signals is responsive to said measured speech activity level.

6. The method of claim 1 wherein said step coding said frame comprises:

generating for said frame a variable number of linear prediction coefficients wherein said variable number of said linear prediction coefficients is responsive to said selected encoding rate;

generating for said frame a variable number of pitch coefficients wherein said variable number of said pitch coefficients is responsive to said selected encoding rate; and

generating for said frame a variable number of codebook excitation values wherein said variable number of said codebook excitation values is responsive to said selected encoding rate.

7. The method of claim 1 wherein said step of determining a level of speech activity comprises summing the squares of the values of said digitized speech samples.

8. The method of claim 7 further comprising the step of generating error protection bits for said data packet.

9. The method of claim 8 wherein said step of generating error protection bits for said data packet wherein the number of said protection bits is responsive to said frame of speech activity level.

10. The method of claim 1 wherein said step of adaptively adjusting speech activity threshold levels comprises the steps of:

comparing said measured speech activity to said at least one of speech activity thresholds and incrementally increasing said at least one of speech activity thresholds toward the level of said frame speech activity when said frame speech activity exceeds said at least one of said speech activity thresholds; and

comparing said measured speech activity to said at least one of speech activity thresholds and decreasing said at least one of speech activity thresholds to the level of said frame speech activity when said frame speech activity is less than said at least one of speech activity thresholds.

11. The method of claim 10 wherein said step of selecting an encoding rate is responsive to an external rate signal.

12. The method of claim 8 wherein said step of generating error protection for said data packet further comprises determining the values of said error protection bits in accordance with a cyclic block code.

13. The method of claim 1 further comprising the step of pre-multiplying said digitized speech samples by a predetermined windowing function.

14. The method of claim 1 further comprising the step of converting said LPC coefficients to line spectral pair (LSP) values.

15. The apparatus of claim 1 wherein said input frame of digitized samples comprises digitized values for approximately twenty milliseconds of speech.

16. The apparatus of claim 1 wherein said input frame of digitized samples comprises approximately 160 digitized samples.

17. The apparatus of claim 1 wherein said output data packet comprises:

one hundred and seventy one bits comprised of forty bits for LPC data, forty bits for pitch data, eighty bits for excitation vector data and eleven bits for error protection when said output data rate is full rate;

eighty bits comprised of twenty bits for LPC information, twenty bits for pitch information and forty bits for excitation vector data when said output data rate is half rate;

forty bits comprised of ten bits for LPC information, ten bits for pitch information and twenty bits for excitation vector data when said output data rate is quarter rate; and

sixteen bits comprised of ten bits for LPC information and six bits for excitation vector information when said output data rate is eighth rate.

18. An apparatus for compressing an acoustical signal into variable rate data comprising:

means for determining a level of audio activity for an input frame of digitized samples of said acoustical signal;

means for selecting an output data rate from a predetermined set of rates based upon said determined level of audio activity within said frame;

means for coding said frame according to a coding format of a set of coding formats for said selected rate to provide a plurality of parameter signals wherein each rate has a corresponding different coding format with each coding format providing a different plurality of parameter signals representing said digitized speech samples in accordance with a speech model; and

means for providing for said frame a corresponding data packet at a data rate corresponding to said selected rate.

19. The apparatus for compressing an acoustical signal of claim 18 wherein said output data packet comprises:

a variable number of bits to represent LPC vector signals of said frame of digitized speech samples, wherein said variable number of bits for representing said LPC vector signals is responsive to said level of audio activity;

a variable number of bits to represent pitch vector signals of said frame of digitized speech samples, wherein said variable number of bits for representing said pitch vector signals is responsive to said level of audio activity; and

a variable number of bits to represent codebook excitation vector signals of said frame of digitized speech samples, wherein said variable number of bits for representing said codebook excitation vector signals is responsive to said level of audio activity.

20. The apparatus for compressing an acoustical signal of claim 18 wherein said means for determining said level of audio activity comprises:

means for determining an energy value for said input frame;

means for comparing said input frame energy with said at least one audio activity thresholds; and

means for providing an indication when said input frame activity exceeds each corresponding one of said at least one audio activity thresholds.

21. The apparatus of claim 20 further comprising a means for adaptively adjusting said at least one of said at least one audio activity thresholds.

22. The apparatus of claim 18 wherein said means for determining said energy of said input frame comprises:

squaring means for squaring said digitized audio samples of a frame; and

summing means for summing said squares of digitized audio samples of a frame.

23. The apparatus of claim 18 wherein said means for determining a level of audio activity comprises:

means for calculating a set of linear predictive coefficients for said input frame of digitized samples of said acoustical signals; and

means for determining said level of audio activity in accordance with at least one of said linear predictive coefficients.

24. The apparatus of claim 18 further comprising means for providing error protection bits for said data packet responsive to said selected output data rate.

25. The apparatus of claim 24 wherein said means for providing error protection bits provides the values of said error protection bits in accordance with a cyclic block code.

26. The apparatus of claim 18 further comprising a means for converting said LPC coefficients to line spectral pair (LSP) values.

27. The apparatus of claim 18 wherein said set of rates comprises full rate, half rate, quarter rate and eighth rate:

28. The apparatus of claim 18 wherein said set of rates comprises 16 Kbps, 8 Kbps, 4 Kbps and 2 Kbps.

29. A circuit for compressing an acoustical signal into variable rate data comprising:

a circuit for determining a level of audio activity for an input frame of digitized samples of said acoustical signal;

a circuit for selecting an output data rate from a predetermined set of rates based upon said determined level of audio activity within said frame;

a circuit for coding said frame according to a coding format of a set of coding formats for said selected rate to provide a plurality of parameter signals wherein each rate has a corresponding different coding format with each coding format providing a different plurality of parameter signals representing said digitized speech samples in accordance with a speech model; and

a circuit for providing for said frame a corresponding data packet at a data rate corresponding to said selected rate.

30. The circuit of claim 29 wherein said circuit for determining said level of audio activity comprises:

a circuit for determining an energy value for said input frame;

a circuit for comparing said input frame energy with said at least one audio activity thresholds; and

a circuit for providing an indication when said input frame activity exceeds each corresponding one of said at least one audio activity thresholds.

31. The circuit of claim 30 wherein said circuit for determining said level of audio activity further comprises a circuit for adaptively adjusting said at least one of said at least one audio activity thresholds.

32. The circuit of claim 29 wherein said circuit for determining said energy of said input frame determines said energy by summing the squares of the values of said digitized samples.

33. The circuit of claim 29 wherein said circuit for determining a level of audio activity determines said energy by calculating a set of linear predictive coefficients for said input frame and determines said level of audio activity in accordance with at least one of said linear predictive coefficients.

34. The circuit of claim 29 wherein said input frame of digitized samples comprises digitized speech for a duration of approximately twenty milliseconds.

35. The circuit of claim 29 wherein said input frame of digitized samples comprises 160 digitized samples.

36. The circuit of claim 29 further comprising means for providing error protection bits for said data packet responsive to said selected output data rate.

37. The circuit of claim 36 further comprises a means for determining the values of said error protection bits in accordance with a cyclic block code.

38. The circuit of claim 37 wherein said cyclic block code operates in accordance with a generator polynomial of 1+x³ +x⁵ +x⁶ +x⁸ +x⁹ +x¹⁰.

39. The circuit of claim 29 further comprising means for pre-multiplying said digitized samples by a predetermined windowing function.

40. The circuit claim 39 wherein said predetermined windowing function is a Hamming window.

41. The circuit of claim 29 wherein said set of rates comprises full rate, half rate, quarter rate and eighth rate.

42. The circuit for claim 29 wherein said set of rates comprises 16 Kbps, 8 Kbps, 4 Kbps and 2 Kbps.

43. The circuit of claim 29 wherein said output data packet comprises:

a variable number of bits to represent LPC vector signals of said frame of digitized speech samples, wherein said variable number of bits for representing said LPC vector signals is responsive to said level of audio activity;

a variable number of bits to represent pitch vector signals of said frame of digitized speech samples, wherein said variable number of bits for representing said pitch vector signals is responsive to said level of audio activity; and

a variable number of bits to represent codebook excitation vector signals of said frame of digitized speech samples, wherein said variable number of bits for representing said codebook excitation vector signals is responsive to said level of audio activity.

44. The circuit of claim 43 wherein said output data packet further comprises a variable number of bits for error protection, wherein said variable number of bits for error protection is responsive to said level of audio activity.

45. The circuit of claim 29 wherein said output data packet comprises:

one hundred and seventy one bits comprised of forty bits for LPC data, forty bits for pitch data, eighty bits for excitation vector data and eleven bits for error protection when said output data rate is full rate;

eighty bits comprised of twenty bits for LPC information, twenty bits for pitch information and forty bits for excitation vector data when said output data rate is half rate;

forty bits comprised of ten bits for LPC information, ten bits for pitch information and twenty bits for excitation vector data when said output data rate is quarter rate; and

sixteen bits comprised of ten bits for LPC information and six bits for excitation vector information when said output data rate is eighth rate.

46. The circuit of claim 29 wherein said means for selecting an encoding rate is responsive to an external rate signal.

47. The circuit of claim 29 further comprising a means for converting said LPC coefficients to line spectral pair (LSP) values.

48. A method of speech signal compression by variable rate coding of frames of digitized speech samples comprising the steps of:

multiplying one frame of digitized speech samples in a sequence of said frames of digitized speech samples by a windowing function to provide a windowed frame of speech data;

calculating a set of autocorrelation coefficients from said windowed frame of speech;

determining an encoding rate from said set of autocorrelation coefficients;

calculating from said set of autocorrelation coefficients a set of linear predictive coding (LPC) coefficients;

converting said set of LPC coefficients to a set of line spectral pair values;

quantizing said set of line spectral pair (LSP) coefficients in accordance with said rate command and said encoding rate;

selecting a pitch value from a predetermined set of pitch values to provide a selected pitch value for each pitch subframe in each frame of digitized speech;

quantizing said selected pitch value in accordance with said encoding rate and said rate command;

selecting a codebook value from a predetermined set of pitch values to provide a selected pitch value for a pitch frame;

quantizing said selected codebook value in accordance with said encoding rate and said rate command; and

generating an output data packet comprising said quantized line spectral pair values, quantized selected pitch value, and quantized selected codebook value.