Method and apparatus for characterization and reconstruction of speech excitation waveforms

• US 5,602,959 A
• Filed: 12/05/1994
• Issued: 02/11/1997
• Est. Priority Date: 12/05/1994
• Status: Expired due to Fees

• Current Assignee: Torsal Technology Group Limited LLC (Intellectual Ventures LLC)
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Jaskie, Cynthia A., Bergstrom, Chad S., Fette, Bruce A., You, Sean S., Wood, Clifford
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): CHOWDHURY, INDRINAL
• Application Number:
  US08/349,638
• Time in Patent Office: 799 Days
• Field of Search: 395/2.12, 395/2.13, 395/2.14, 395/2.15, 395/2.16, 395/2.28
• US Class Current: 704/205
• CPC Class Codes: G10L 19/02 using spectral analysis, e.g. transform vocoders or subband vocoders
• CPC Class Codes: G10L 19/10 the excitation function being a multipulse excitation
• CPC Class Codes: G10L 19/125 Pitch excitation, e.g. pitch synchronous innovation CELP [PSI-CELP]
• CPC Class Codes: G10L 2019/0011 Long term prediction filters, i.e. pitch estimation

Abstract

A vocoder device and corresponding method characterizes and reconstructs speech excitation. An excitation analysis portion performs a cyclic excitation transformation process on a target excitation segment by rotating a peak amplitude to a beginning buffer location. The excitation phase representation is dealiased using multiple dealiasing passes based on the phase slope variance. Both primary and secondary excitation components are characterized, where the secondary excitation is characterized based on a computation of the error between the characterized primary excitation and the original excitation. Alternatively, an excitation pulse compression filter is applied to the target, resulting in a symmetric target. The symmetric target is characterized by normalizing half the symmetric target. The synthesis portion performs reconstruction and synthesis of the characterized excitation based on the characterization method employed by the analysis portion.

Claims

1. A method of encoding speech comprising the steps of:

a) performing linear predictive coding (LPC) on a plurality of digital speech samples to obtain an excitation waveform;

b) selecting a target excitation segment from the excitation waveform, wherein the target excitation segment is selected to be synchronous to a pitch of the excitation waveform;

c) performing a cyclic excitation transformation of the target excitation segment by placing the target excitation segment in a buffer, cyclically shifting the target excitation segment until a peak of the target excitation segment is positioned at a beginning of the buffer, and performing a time-domain to frequency-domain transformation of the target excitation segment, wherein a result of the cyclic excitation transformation is a transformed target excitation segment;

d) characterizing the transformed target excitation segment, resulting in a characterized excitation waveform;

e) generating characterized, encoded excitation by post-processing the characterized excitation waveform; and

f) storing a bitstream that incorporates the characterized, encoded excitation.

2. The method as claimed in claim 1, further comprising the step of weighting the target excitation segment prior to performing the step of performing the cyclic excitation transformation.

3. The method as claimed in claim 1, further comprising the step of transmitting the bitstream.

4. A method of encoding speech comprising the steps of:

a) performing linear predictive coding (LPC) on a plurality of digital speech samples to obtain an excitation waveform;

b) selecting a target excitation segment from the excitation waveform, wherein the target excitation segment is selected to be synchronous to a pitch of the excitation waveform;

c) generating a characterized excitation waveform from the target excitation segment by performing excitation pulse compression filtering and characterizing a symmetric excitation, wherein performing the excitation pulse compression filtering comprises the steps of;

c1) determining matched filter coefficients that serve to cancel group delay characteristics of the target excitation segment, and

c2) applying a matched filter defined by the matched filter coefficients to the target excitation segment, resulting in a symmetric excitation;

d) generating characterized, encoded excitation by post-processing the characterized excitation waveform; and

e) storing a bitstream that incorporates the characterized, encoded excitation.

5. The method as claimed in claim 4, further comprising the step of transmitting the bitstream.

6. A speech vocoder analysis device comprising:

a memory device for storing digital speech samples;

an analysis processor coupled to the memory device for generating an excitation waveform by;

performing LPC analysis on a plurality of digital speech samples,

selecting a target excitation segment from the excitation waveform, wherein the target excitation segment is selected to be synchronous to a pitch of the excitation waveform,

performing a cyclic excitation transformation of the target excitation segment by placing the target excitation segment in a buffer, cyclically shifting the target excitation segment until a peak of the target excitation segment is positioned at a beginning of the buffer, and performing a time-domain to frequency-domain transformation of the target excitation segment, wherein a result of the cyclic excitation transformation is a transformed target excitation segment,

characterizing the transformed target excitation segment, resulting in a characterized excitation waveform,

generating characterized, encoded excitation by post-processing the characterized excitation waveform, and

storing a bitstream that incorporates the characterized, encoded excitation; and

a modem coupled to the analysis processor.

7. A speech vocoder analysis device comprising:

an analog-to-digital converter for converting input speech signals into digital speech samples;

an analysis processor coupled to the memory device for generating an excitation waveform by;

performing LPC analysis on a plurality of digital speech samples,

selecting a target excitation segment from the excitation waveform, wherein the target excitation segment is selected to be synchronous to a pitch of the excitation waveform,

performing a cyclic excitation transformation of the target excitation segment by placing the target excitation segment in a buffer, cyclically shifting the target excitation segment until a peak of the target excitation segment is positioned at a beginning of the buffer, and performing a time-domain to frequency-domain transformation of the target excitation segment, wherein a result of the cyclic excitation transformation is a transformed target excitation segment,

characterizing the transformed target excitation segment, resulting in a characterized excitation waveform,

generating characterized, encoded excitation by post-processing the characterized excitation waveform, and

storing a bitstream that incorporates the characterized, encoded excitation; and

a modem coupled to the analysis processor.

8. The speech vocoder analysis device as claimed in claim 7 further comprising:

a microphone coupled to the analog-to-digital converter; and

a memory device coupled to the analog-to-digital converter and the analysis processor.

9. A speech vocoder analysis device comprising:

an analog-to-digital converter for converting input speech signals into digital speech samples;

an analysis processor coupled to the analog-to-digital converter for generating an excitation waveform by performing LPC analysis on a plurality of digital speech samples, selecting a target excitation segment from the excitation waveform, generating a characterized excitation waveform by performing excitation pulse compression filtering and characterizing a symmetric excitation of the target excitation segment, wherein performing the excitation pulse compression filtering comprises the steps of;

c1) determining matched filter coefficients that serve to cancel group delay characteristics of the target excitation segment, and

c2) applying a matched filter defined by the matched filter coefficients to the target excitation segment, resulting in a symmetric excitation,

the analysis processor further for generating characterized, encoded excitation by post-processing the characterized excitation waveform, and storing a bitstream that incorporates the characterized, encoded excitation; and

a modem coupled to the analysis processor.