In this thesis, the performances of adaptive differential pulse code modulation (ADPCM) originally proposed by Cummiskey et al. and delayed-decision ADPCM(DD-ADPCM) systems have been studied by computer simulation with particular emphasis on reducing the computational complexity and on improving the performance in a noisy channel.
The DD-ADPCM that yields improved performance especially at low bit rates compared with the conventional ADPCM system showed a signal-to-quantization noise ratio (SQNR) gain of about 1dB per one sample delay. To reduce the computational loads that are caused by increment of the delay samples, a block decision scheme has been proposed for the DD-ADPCM.
In addition, we have studied the use of convolutional coding and the reduced-state Viterbi algorithm in the DD-ADPCM so as to reduce the effect of noisy channel. The reduced-state Viterbi algorithm yielded almost the same performance as the conventional Viterbi algorithm, while reducing the computational load significantly.
본 논문에서는 cummiskey 등에 의해 제안된 ADPCM system 과 DD-ADPCM system 의 performance 에 관한 연구가 계산량의 감소와 noisy channel 특성의 개선에 중점을 두고 computer simulation 을 통하여 수행되었다.
DD-ADPCM은 ADPCM에 비해 낮은 bit rate 에서는 매 sample delay 의 증가에 따라 1 dB 정도의 SONR 증가를 보여주고 있다. Delay sample 의 증가에 따른 계산량의 증가를 줄이기 위하여, block decision 방식을 고려하였다.
또한 noisy channel의 영향을 줄이기 위하여 DD-ADPCM에 convolutional coding 과 reduced-state VA을 사용하였다. VA와 거의 비슷한 특성을 보이는 reduced-state VA 은 계산량 감소에 관하여 커다란 효과를 나타내고 있다.