Abstract: The speech formant frequency estimation is done based on piecewise linear predictive coding.Multi-channel filter banks are applied to decomposing the speech spectrum.Suitable inverse filters are chose to approach the STFT of different spectral bands.The speech formant frequency is obtained by the inverse filters later.The results show that，compared with the conventional formant estimation methods，the method based on piecewise linear predictive coding improves the resolution and veracity of the speech formant frequency estimation，and is not sensitive to noise.

Key words: speech formant frequency, piecewise linear predictive coding, filter banks

摘要： 基于分段线性预测算法估计语音的共振峰频率，运用多通道的滤波器组对语音的频段进行划分，然后选择合适的逆滤波器逼近不同频段的短时频谱，最后依据该逆滤波器估计共振峰频率。实验结果表明，与传统方法相比，该方法提高了语音共振峰频率估计时的分辨率与准确性，受噪声的影响较小。

关键词: 共振峰频率, 分段线性预测, 滤波器组