摘要
该文将小波变换应用于具有连续语音特征的三字词语音的基频提取,并针对实验中出现的问题对算法进行了改进,提出了一种新的基于小波变换的语音基频检测算法。该算法主要包括:离散小波变换计算、基于投票策略的基频点选择和基频起点确定、基频检查、异常点修正、头尾漏点处理以及基于投票策略的基频点精确定位。实验表明,该算法较好地克服了基于小波变换传统算法的不足,更适合于连续语音的基频提取,缺陷是需要较多的计算时间,不太适合于实时性要求较高的系统。
In this paper wavelet transform is applied to pitch detection of Chinese trisyllabic word with characteristic of continuous speech, a new algorithm for pitch detection of speech signals using wavelet transform is proposed for the question appearing in pitch extraction. It is made up of the following processes: calculating discrete wavelet transform, selecting pitch points and determining the first pitch point based on voting strategy, checking pitch, modifying abnormal pitch point, processing pitch miss point at the beginning and end of a frame of speech signals, relocation of pitch point based on voting strategy. It has been found that this algorithm is valid, and very suits to process continuous speech. But it needs much time for computation, so it is not suit for real-time system.
出处
《电子与信息学报》
EI
CSCD
北大核心
2002年第9期1218-1224,共7页
Journal of Electronics & Information Technology
基金
博士点基金 (98035904)
