This paper presents a Waveform Interpolation (WI) speech coder, which characteristic waveform extraction rate is adaptive to the feature of the input frame. Efficient pitch estimation algorithm is based on the principle of maximizing double weighted Long Time Prediction(LTP) gain and uses the forward pitch detection. The waveform extraction rate and the update rate of SEW(Sulocoly Evoloiuy Waveform) and REW(Rapidly Evolving Waveform) are based on the three features: pitch cycle, voicing degree and stationary degree of waveform surface. Tests indicate that the proposed WI coding algorithm has lower average bit rate and computing complexity compared to the fixed-extraction-rate WI coder and obviously deliver better quality than FS1016 CELP at 4.8kbps.
[1] 鲍长春. 低比特率数字语音编码基础[M]. 北京:北京工业大学出版社,2001: 233-234. [2] Kleijin W B. Continuous representation in linear predictive coding[A]. IEEE ICASSP91[C]. Toronto, 1991, 201-204. [3] Kleijin W B and Haagen J. Transformation and decomposition of the speech signal for coding[J].IEEE Signal Processing Letters.1994, 1(9):136-139 [4] Kleijin W B. A speech coder based on decomposition of characteristic waveforms[A]. IEEE ICASSP95[C]. Detroit, 1995: 508-511. [5] Kleijing W B.[J].Shoham Y, and Sen D, et al.. A low-complexity waveform interpolation coder[A]. IEEE ICASSP96[C]. Atlanta.1996,:- [6] 鲍长春,樊昌信. 基于归一化互相关函数的基音检测算法[J].通信学报,1998, 19(10): 27-31. [7] [7] 3GPP TS 26.190. AMR Wideband speech codec; Transcoding functions[S], 2001, Release 5. [8] Das A, Rao A V, and Gersho A. Variable-dimension vector quantizaiton[J].IEEE Signal Processing Letters.1996, 3(7):200-202