摘要
適應性差分脈碼調變(Adaptive differential pulse code modulation, ADPCM) 是目 前常見的一種語音編碼方法，具有良好品質，適當的編碼率，以及低雜度低的特 性。ADPCM有很多種變形，其中Microsoft ADPCM (MS-ADPCM)是微軟視窗系 統建議的一種編碼方式。 根據Shin及Kim的研究[4]，使用以合成作分析(Analysis by Synthesis, ABS) 的方法，可以進一步改善ADPCM語音重建訊號的品質。在我們先前的研究當 中[5]，曾提出一個以合成作分析架構下的ADPCM編碼方法，使用該方法可以得 到比Shin及Kim所提的方法有更好的編碼效能。 在這篇論文中，我們將ABS的概念應用在MS-ADPCM之上。MS-ADPCM是 以區塊為單位的編碼方法，每個區塊都有一個描述檔頭，其後則緊跟著一系 列的四位元組(nibble)編碼符元，因此我們將所提的方法[5]做適當修改，以符 合MS-ADPCM的區塊編碼架構。我們提出一些實際的編碼方法，並探討其編碼 效能。我們從TIMIT語音資料庫中取出數個語音檔，使用我們所提的方法和原始 的MS-ADPCM方法做比較，根據我們的模擬結果，其編碼效能可以大幅改善，有 些甚至可以高達10 dB以上且PESQ亦可提升0.4以上。
並列摘要
Adaptive differential pulse code modulation (ADPCM) is a widely used speech codec because of its simplicity, good performance, and moderate bit rates. There are various flavors of ADPCM, and Microsoft ADPCM (MS-ADPCM) is the algorithm suggested by Microsoft. According to the study of Shin and Kim [4], the coding performance of ADPCM can be improved by using analysis by synthesis (ABS) technique. In our previous study [5], a general encoding method which comines ADPCM and ABS concept is proposed and has shown to have better coding performance than Shin and Kim’s algorithm. In this thesis, we apply the ABS technique to MS-ADPCM. MS-ADPCM is organized in blocks. Each block has a preamble and a series of coded ADPCM nibbles. Hence we modify the algorithm proposed in [5] to suit the block-based structure in MS-ADPCM. We study the performance limits and propose some practical algorithms. According to our simulations, by using the speech samples from TIMIT, the proposed algorithms can improve the output signal-to-noise ratio (SNR) up to 10 dB and PESQ(Perceptual evaluation of speech quality) up to 0.4 when compared to the original MS-ADPCM algorithm.