用模拟退火算法实现语音识别中的矢量量化
ANNEALING VECTOR QUANIZATION IN SPEECH RECOGNITION
-
摘要: 矢量量化在语音识别中有着重要的作用。经典的K均值算法收敛速度快,但极易收敛于局部最佳点;其它的一系列改进算法在克服其局部收敛问题的同时,又显著增加了运算量。本文提出了用模拟退火算法实现语音识别中的矢量量化过程,能够较好地协调运算量和收敛质量之间的矛盾。文章讨论了具体算法,并给出了实验数据。结果表明该方法的综合性能优于现有算法,具有较高的实用价值。Abstract: Vector quantization plays an important role in speech recognition.Traditional K-means algorithm owns the advantage of fast convergence, but it is difficult to get the global optimal result.Some modified algorithms have been proposed to overcome this drawback,but they also increase the computation greatly.In this papsr,a new algorithm which is based on annealing algorithm is proposed to compromise the contradiction.In the rest of the paper,the details of the algorithm and related experiments are given.The results demonstrate the algorithm is more effective than other methods.
-
Rabiner L.R.A tutorial on hidden Markov models and selected applications in speech recognition．Proc．IEEE，1989、77(22)：257-285[2]Huo Q. Contextual vector quantization for speech recognition with discrete hidden Markov model．Pattern Recognition，1995、28(4)：513-517[3]胡光锐．语音处理与识别．上海；上海科学技术文献出版社，1994，200-297．[4]陈尚勤，等．近代语音识别．成都；电子科技大学出版社，1991：66-85．[5]Press W．H．等著，傅祖芸，等译．C语言数值算法程序大全．北京:电子工业出版社，1995，378-389．[6]Tourneret J．Nomality of a non-linear transformation of AR parameters：Application to reflection and cepstrum coefficients．Signal Processing，1997、62(1)：l-14[7]Tohkura YA．Weighted ceptral distance measure for speech recognition．IEEE Trans．On ASSP，1987 ASSP-35(10)：1414-1422．[8]马卡尔著，娄乃英译．语音信号线性预测．北京：中国铁道出版社，1987，第1章． -
计量
- 文章访问数: 2125
- HTML全文浏览量: 152
- PDF下载量: 455
- 被引次数: 0
下载:
