一种基于倒谱法的基音周期检测改进算法

Improved algorithm of pitch detection based on cepstrum

通过对常规语音信号处理分析,设计一种基于倒谱法的基音周期检测改进算法。其过程是先对语音信号进行预处理、去趋势项处理及去噪声处理,,然后通过语音分析中的线性化处理、伯格算法和中值滤波算法对预处理过的语音信号进一步优化,并利用仿真软件验证改进增强算法,改进算法可有效的减少外界环境及共振峰等因素的干扰,相对于常规的语音基音检测算法其鲁棒性、稳定性更强。

By analyzing the detection of voice speech signal, carry out pitch detection using improved algorithm based on cep- strum. Firstly, deal with speech signal removing linear trend and reduce noise by spectral subtraction method based on multi- taper estimation, then enhance preprocessed speech signal by, then enhance the voice by linear prediction coding （LPC）, Burg algorithm and smoothing algorithm. Verify and improve algorithm via the simulation experiments. The experimental re- suits indicate that this algorithm an reduce trouble from external environment and resonance peak etc, and is more robust and steady than conventional pitch detection methods.