波形編碼是在時域上 (time-domain) 針對訊號進行編碼的一種編碼技術,由於其編碼簡單以及性能良好而被使用。常見的波形編碼,如 PCM 或 ADPCM 等。以合成作分析 (Analysis by Synthesis, ABS) 是另一種編碼技術,主要常見於以模型為基礎 (model-based) 的語音編碼技術。主要的訴求是透過合成的方法,在許多可能的編碼參數中,尋找較為合適的參數。2006年,Shin及Kim提出一種方法[2],可以使用ABS的概念來改善ADPCM的編碼性能。之後不久,Yang及Wang提出另一種方法[4],同樣使用ABS的概念,但可以得到更好的編碼性能。根據上述的研究,可以知道波形編碼可以透過ABS的概念來改善其編碼性能,而能達到的最佳性能主要是由其解碼部份決定。本論文針對幾種常見的波形編碼,包括 ADM, CVSD, IMA ADPCM, Dialogic ADPCM等,依循[4]的方法和設計準則,加入ABS的概念,探討其最佳的編碼性能。同時也利用簡化的搜尋方法,在適當的複雜度及較佳的性能中取得平衡。我們建立了許多數據表格,希望這些結果可以供未來的研究或應用做為參考。
Waveform coding is a coding technique in the time domain for the signal coding. It is commonly used due to its simplicity and good performance. Common waveform coding includes PCM, ADPCM, etc. Analysis by Synthesis (ABS) is another coding technique which is commonly used in model-based speech coding. The key is to choose a good parameter combination from large possible candidates by using synthesis method. In 2006, Shin and Kim [2] proposed a method to improve the coding performance of ADPCM by using ABS concept. Later, Yang and Wang [4] proposed another method to apply ABS in ADPCM with further improvement. According to the previous study, waveform coding can be improved by ABS concept, and its extreme gain is determined by the decoding part. In this thesis, we follow the methods and guides of [4] to apply the ABS method in a variety of common waveform codings, including ADM, CVSD, IMA ADPCM, and Dialogic ADPCM. We are intended to explore the best performances of these waveform codings and also try to find a simplified searching method which gives a balance in the appropriate complexity and better performance. We compile several tables and hope these results can be used as a reference for further researches or applications.