摘要
基于对语音信号的分析,借鉴频率分带技术和码激励线性预测编码中传输残差信号可提高合成语音鲁棒性的思想,建立了双带激励模型,同时对基音检测、矢量量化、语音合成等环节加以改进,提出了最高码率为2400bit/s,最低码率为80bit/s的变码率增强型双带激励LPC(LinearPredictiveCoding)低速语音编码算法。仿真结果表明:该算法在有效降低平均码率的前提下能保证较高的合成语音质量。
Based on the analysis of speech signals, using sub-band division technology and the idea of transmitting residual signals in CELP(Code Excited Linear Predi ctive) coding algorithm, a variable-rate enhanced double-band excitation LPC(Linear Predictive Coding) low bit rate speech coding algorithm is presente d. In order t o improve the robustness and quality of the synthesis speech, some new methods an d ameliorations are employed in pitch detection, vector quantization, and speech synthesis. Simulation results indicate that this algorithm ensures good quality of synthesis speech with low bits.
出处
《吉林大学学报(信息科学版)》
CAS
2003年第2期110-116,共7页
Journal of Jilin University(Information Science Edition)
基金
吉林省高科技开发基金资助项目(20010316)
关键词
双带激励
低速语音编码
变码率
频率分带
Double-band excitation
Low bit speech coding
Vari able-rate
Sub-band classification