(A) performance study of speech bandwidth compression systems음성 대역폭 축소 시스템의 성능에 관한 연구

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 487
  • Download : 0
In this dissertation, the performances of speech bandwidth compression systems have been studied, and new results have been obtained in four areas that attract recent research interests. The speech bandwidth compression systems chosen for performance study are pulse code modulation (PCM), differential PCM(DPCM), delta modulation (DM), their adaptive versions (i.e., adaptive PCM(APCM), adaptive DPCM (ADPCM) and adaptive DM(ADM), and the generalized residual-excited linear predictive (RELP) coder. First, the performances of APCM, ADPCM and ADM systems have been compared at the bit rate of 32 kbits/s using conventional signal-to-quantization noise ratio (SQNR) and segmented SQNR($SQNR_{SEG}$) as performance criteria. In APCM and ADPCM, both the instantaneously and syllabically companding methods are applied to each uniform or nonuniform quantizer. For ADM systems, three representative ADM systems have been considered. They are continuously variable slope DM(CVSD), constant factor DM (CFDM) and hybrid companding DM(HCDM). Among these waveform coders, the performance of HDCM is superior to other coders. The performance cross-over between ADPCM and ADM systems occurs at the bit rate slightly above 32 kbits/s. Therefore, the use of ADM systems (particularly HCDM system) at the transmission rate of 32kbits/s or below is preferred to ADPCM. Second, analytical studies on the performances of waveform coders have been done. New results for the frequency domain modeling of waveform coders have been obtained. These models are obtained based on the overload characteristics of each waveform coder in the frequency domain. The overload characteristics are determined by two threshold levels which can decide the coder in one of three operating regions: overload region, granular region and the region below the threshold of coding. The performances of waveform coders are then predicted by using their corresponding frequency domain models with real speech. Close agreements between ...
Advisors
Un, Chong-Kwanresearcher은종관researcher
Description
한국과학기술원 : 전기 및 전자공학과,
Publisher
한국과학기술원
Issue Date
1983
Identifier
60851/325007 / 000765106
Language
eng
Description

학위논문(박사) - 한국과학기술원 : 전기 및 전자공학과, 1983.2, [ xiii, 217 p. ]

URI
http://hdl.handle.net/10203/35737
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=60851&flag=dissertation
Appears in Collection
EE-Theses_Ph.D.(박사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0