摘要
针对目前使用的时域语音时长调整算法计算量大、难以保证实时性的问题,深入研究语音时长调整算法的原理,并结合语音自身的短时平稳性、准周期特性和信号的频率特性,提出三种解决途径(只搜索叠加部分原则、隔点搜索原则、隔点计算相似性原则),在保证语音质量不降低的同时,大幅减少冗余的计算量,实验结果表明,该方法调整后的语音质量高、计算速度快,通过与原始算法的对比证明了以上结论,该方法能广泛应用于实际中,尤其是应用于实时性要求较高的场合。
Three approaches to solving the problem of the large computation amounts of the time-scale modification algorithm have been proposed through deeply studying the principle of the algorithm and combining with the short time stationary,quasi-periodic and frequency characteristics of the speech signal.The new proposed fast algorithm could reduce the computation significantly,and meantime maintain the quality of speech.The experimental results show that the modified speech could maintain the quality,and the new algorithm increases the speed of computation and can be widely used in practice.
出处
《声学技术》
CSCD
2010年第5期507-511,共5页
Technical Acoustics
基金
广西科学基金资助项目(0639028)
关键词
语音
时长调整
快速算法
speech
time-scale modification
fast algorithm
speed and quality
