Variable rate speech coding based on subband measures of spectral flatness /
The development of digital multiple-access schemes for telephony and personal communications networks, in addition to the rapid standardization of digital audio-video delivery techniques, has placed a renewed emphasis on efficient compression of speech. Code-excited linear prediction (CELP) is the...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1995.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=742535101&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | The development of digital multiple-access schemes for telephony and personal communications networks, in addition to the rapid standardization of digital audio-video delivery techniques, has placed a renewed emphasis on efficient compression of speech. Code-excited linear prediction (CELP) is the predominant coding methodology for medium-rate speech coding due to the adoption of several international and U.S. fixed-rate standards. With the standardization of a CELP- based variable-rate coder for digital spread-spectrum telephony, many variable-rate techniques which had not been previously studied have become an area of significant research interest. Variable-rate coding of speech is composed of two distinct components: voice-activity detection (VAD) and coding of active speech. In this research, we focus on new techniques to perform VAD and to efficiently encode time- varying parameters of active speech. These techniques are based on subband measures of spectral shape. This approach produces a departure from the usual methods used for rate allocation which are based on (subband) energy. Here, we use the concept of spectral flatness to modulate the allocation of rate and to determine coding procedures. We also make use of some highly speech-specific redundancy (previously untested in standard approaches) to efficiently encode spectral parameters. We test these techniques in the context of a multi-mode, variable-rate CELP coder and demonstrate that with appropriate assumptions regarding voice-activity, our CELP coder can maintain communications quality in the encoded speech at an average rate below 2000 bits/sec. |
|---|---|
| Item Description: | Vita. "Major Subject: Electrical Engineering". |
| Physical Description: | xii, 137 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilms Inc. |
| Bibliography: | Includes bibliographical references. |