US5247579A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Methods for speech transmission
Assignee: Digital Voice Systems Inc
•
Filing: 1991-12-03
•
Granted: 1993-09-21
•
Expires: 2011-12-03
Abstract
Abstract The performance of speech coding in the presence of bit errors is improved. The quantized parameter bits are grouped into several categories according to their sensitivity to bit errors. More effective error correction codes are used to encode the most sensitive parameter bits, while less effective error correction codes are used to encode the less sensitive parameter bits. This method improves the efficiency of the error correction and improves the performance if the total bit rate is limited. The perceived quality of coded speech is improved. A smoothed spectral envelope is created in the frequency domain. The ratio between the actual spectral envelope and the smoothed spectral envelope is used to enhance the spectral envelope. This reduces distortion which is contained in the spectral envelope.
Inventors
- 👤 John C. Hardwick
- 👤 Jae S. Lim
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-25 06:22
CPC Classifications
G; G10; G10L; G10L19/00; G10L19/04; G10L19/08; G10L19/087; G10L19/005; G10L19/02; G10L19/032; G10L19/038; G10L21/00; G10L21/02; G10L21/0316; G10L21/0364; H; H03; H03M; H03M13/00; H03M13/35; G10L19/06; G10L19/10; G10L21/0208; G10L21/0216; G10L21/0232
Claims Analysis
Total Claims: 22
Claim 1
Independent Claim
140 words
A method of encoding speech wherein the speech is encoded using a speech model characterized by model parameters, wherein the speech is broken into time segments and for each segment model parameters are estimated and quantized, and wherein at least some of the quantized model parameters are coded using error correction coding, said method comprising the steps of: using a first type of error correction coding to code the quantized model parameters in a first group, using a second type of error correction coding to code the quantized model parameters in a second group, the first group containing quantized model parameters that are more sensitive to bit errors than are the quantized model parameters in the second group, the first type of error correction coding adding a greater number of additional bits than the second type of error correction coding.
Claim 2
Independent Claim
140 words
A method of encoding speech wherein the speech is encoded using a speech model characterized by model parameters, wherein the speech is broken into time segments and for each segment model parameters are estimated and quantized, and wherein at least some of the quantized model parameters are coded using error correction coding, said method comprising the steps of: using a first type of error correction coding to code the quantized model parameters in a first group, using a second type of error correction coding to code the quantized model parameters in a second group, the first group containing quantized model parameters that are more sensitive to bit errors than are the quantized model parameters in the second group, the first type of error correction coding adding a greater number of additional bits than the second type of error correction coding.
Claim 3
Dependent Claim
22 words
The method of claim 1 wherein the first and second types of error correction coding are Golay codes and Hamming codes, respectively.
Claim 4
Dependent Claim
25 words
The method of claim 1 or 2 wherein soft decision decoding is employed in decoding at least one of the groups of quantized model parameters.
Claim 5
Dependent Claim
17 words
The method of claim 1 or 2 wherein error rates are estimated using the error correction codes.
Claim 6
Dependent Claim
30 words
The method of claim 4 wherein one or more model parameters are smoothed across a plurality of segments based on estimated error rates, and wherein said model parameters have values.
Claim 7
Dependent Claim
13 words
The method of claim 5 wherein the model parameters smoothed include voiced/unvoiced decisions.
Claim 8
Dependent Claim
38 words
The method of claim 5 wherein the model parameters smoothed include parameters for at least one of the following speech coders: Multi-Band Excitation (MBE) speech coder, Improved Multi-Band Excitation (IMBE) speech coder, or sinusoidal transform speech coder (STC).
Claim 9
Dependent Claim
36 words
The method of claim 5 wherein the value of one or more model parameters in a previous segment is repeated in a current segment when the estimated error rate for the parameters exceeds a predetermined level.
Claim 10
Independent Claim
107 words
A method of encoding speech wherein the speech is encoded using a speech model characterized by model parameters, wherein the speech is broken into time segments and for each segment model parameters are estimated and quantized, and wherein at least some of the quantized model parameters are coded using error correction coding, and wherein speech is synthesized from decoded quantized model parameters, said method comprising the steps of: using the error correction coding during synthesis to estimate an error rate, repeating in a current segment one or more model parameters from a previous segment when the error rate for the quantized model parameters exceeds a predetermined level.
Claim 11
Independent Claim
107 words
A method of encoding speech wherein the speech is encoded using a speech model characterized by model parameters, wherein the speech is broken into time segments and for each segment model parameters are estimated and quantized, and wherein at least some of the quantized model parameters are coded using error correction coding, and wherein speech is synthesized from decoded quantized model parameters, said method comprising the steps of: using the error correction coding during synthesis to estimate an error rate, repeating in a current segment one or more model parameters from a previous segment when the error rate for the quantized model parameters exceeds a predetermined level.
Claim 12
Dependent Claim
66 words
The method of claim 1, 2 or 9 wherein the quantized speech model parameters are those associated with at least one of the following speech coders: a Multi-Band Excitation (MBE) speech coder, an Improved Multi-Band Excitation (IMBE) speech coder, a linear prediction speech coder, a channel speech coder, a homomorphic speech coder, a sinusoidal transform speech coder (STC), and a code-excited linear prediction (CELP) speech coder.
Claim 13
Independent Claim
129 words
A method of decoding speech wherein the speech is decoded and synthesized using a speech model characterized by model parameters, wherein the model parameters have been quantized and encoded, and wherein at least some of the quantized model parameters are decoded using error correction decoding, said method comprising the steps of: using a first type of error correction decoding to decode the quantized model parameters in a first group, using a second type of error correction decoding to decode the quantized model parameters in a second group, the first group containing quantized model parameters that are more sensitive to bit errors than are the quantized model parameters in the second group, the first type of decoding utilizing a greater number of additional bits than the second type of decoding.
Claim 14
Independent Claim
129 words
A method of decoding speech wherein the speech is decoded and synthesized using a speech model characterized by model parameters, wherein the model parameters have been quantized and encoded, and wherein at least some of the quantized model parameters are decoded using error correction decoding, said method comprising the steps of: using a first type of error correction decoding to decode the quantized model parameters in a first group, using a second type of error correction decoding to decode the quantized model parameters in a second group, the first group containing quantized model parameters that are more sensitive to bit errors than are the quantized model parameters in the second group, the first type of decoding utilizing a greater number of additional bits than the second type of decoding.
Claim 15
Dependent Claim
24 words
The method of claim 11 wherein the first and second types of error correction decoding are Golay code decoding and Hamming code decoding, respectively.
Claim 16
Dependent Claim
65 words
The method of claim 11 or 12 wherein the quantized speech model parameters are those associated with at least one of the following speech coders: a Multi-Band Excitation (MBE) speech coder, an Improved Multi-Band Excitation (IMBE) speech coder, a linear prediction speech coder, a channel speech coder, a homomorphic speech coder, a sinusoidal transform speech coder (STC), and a code-excited linear prediction (CELP) speech coder.
Claim 17
Independent Claim
163 words
A method of enhancing speech wherein a speech signal is broken into segments, and wherein frequency domain representations of a segment is determined to provide an unsmoothed spectral envelope of the segment, and speech is synthesized from an enhanced spectral envelope, said method comprising the steps of: smoothing the unsmoothed spectral envelope of the segment to generate a smoothed spectral envelope generating the enhanced spectral envelope by increasing the amplitude of the smoothed spectral envelope in at least some frequency regions for which the unsmoothed spectral envelope has greater amplitude than the smoothed spectral envelope, and decreasing the amplitude of the smoothed spectral envelope in at least some frequency regions for which the unsmoothed spectral envelope has lesser amplitude than the smoothed spectral envelope, thereby generating an enhanced spectral envelope in which the amplitudes in some regions are increased above the amplitudes of the smoothed spectral envelope and the amplitudes in other regions are decreased below the amplitudes of the smoothed spectral envelope.
Claim 18
Independent Claim
163 words
A method of enhancing speech wherein a speech signal is broken into segments, and wherein frequency domain representations of a segment is determined to provide an unsmoothed spectral envelope of the segment, and speech is synthesized from an enhanced spectral envelope, said method comprising the steps of: smoothing the unsmoothed spectral envelope of the segment to generate a smoothed spectral envelope generating the enhanced spectral envelope by increasing the amplitude of the smoothed spectral envelope in at least some frequency regions for which the unsmoothed spectral envelope has greater amplitude than the smoothed spectral envelope, and decreasing the amplitude of the smoothed spectral envelope in at least some frequency regions for which the unsmoothed spectral envelope has lesser amplitude than the smoothed spectral envelope, thereby generating an enhanced spectral envelope in which the amplitudes in some regions are increased above the amplitudes of the smoothed spectral envelope and the amplitudes in other regions are decreased below the amplitudes of the smoothed spectral envelope.
Claim 19
Dependent Claim
49 words
The method of claim 14 wherein the frequency domain representation of the unsmoothed spectral envelope is the set of spectral amplitudes forming parameters of at least one of the following speech coders: Multi-Band Excitation (MBE) speech coder, Improved Multi-Band Excitation (IMBE) speech coder, or sinusoidal transform speech coder (STC).
Claim 20
Dependent Claim
24 words
The method of claim 14 or 15 wherein the smoothed spectral envelope is generated by estimating a low-order model from the unsmoothed spectral envelope.
Claim 21
Dependent Claim
25 words
The method of claim 16 wherein the low-order model is one of the following models: an all-pole model, a cepstral model, or a polynomial model.
Claim 22
Dependent Claim
24 words
The method of claim 14 or 15 wherein the smoothed spectral envelope is generated by convolving the unsmoothed spectral envelope with a smoothing function.
References Cited (Backward Citations)
US3982070A
Cited
US4276647A
Cited
US4648091A
Cited
US4744086A
Cited
US5023910A
Cited
US5007094A
Cited
US5036515A
Cited
US5081681A
Cited
US5081681B1
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1996-09-20 | 1997-03-22 | 1997-09-21 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2000-09-20 | 2001-03-22 | 2001-09-21 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2004-09-20 | 2005-03-22 | 2005-09-21 | $7700.00 | Window Closed / Lapsed |