TETRA语音编码中基音预处理算法的优化

本文在TETRA语音编码的预处理部分引入了数值滤波算法。该算法能够有效去除声道的共振峰结构对基音检测的影响，进一步试验采用去均值。低通滤波和数值滤波等算法来替代原有的预处理，可以得到更好的优化效果。我们设计了普通语音、基音有抖动的语音和带有噪声干扰的语音未检验优化算法的性能。在这两种优化算法中都可以发现处理后的语音信号在时域上表现出明显的周期性特征，同时在频域上原始语音的共振峰影响被消除或得到有效抑制。

Optimization of Pitch Preprocessing in TETRA Speech Encoder

An algorithm using numerical value filter is introduced to pre-process original speech signal before inputting it into the speech encoder of TETRA. Using this algorithm the formant effect caused by the track in the process of pitch detection can be removed effectively. Instead of traditional pre-processing in TETRA encoder, another more effective algorithm consisting of average value deduction, low-pass filter and numerical value filter is further proposed. Normal speech, speech with pitch dithering and speech background noise have been used to test the performance of the optimized algorithm, Speech signal after processed by both proposed algorithms presented obvious periodicity in time domain. At the same time, effect of the formant in frequency domain can also be suppressed effectively.