离散时间域的 Gabor 谱分析及其在语音处理中的应用

The Gabor Spectrum Analysis in Discrete Time Domain and Its Application to Speech Signal Processing

把Ｇａｂｏｒ谱应用于离散时间域，对典型信号进行处理和分析，并与短时傅里叶变换和离散Ｗｉｇｎｅｒ分布相比较，证明了离散Ｇａｂｏｒ谱不仅在时频域具有和离散Ｗｉｇｎｅｒ分布相同的高分辨率，优于短时傅里叶变换，而且能有效地消除离散Ｗｉｇｎｅｒ分布中交叉干扰的影响．文中还进一步应用离散Ｇａｂｏｒ谱，对汉语语音信号进行了处理，结果表明它可以较好地描述语音信号频谱的时变特性，为语音识别和话者识别开创了一种新的途径．

This paper applied the Gabor spectrum to discrete time domain. The discrete Gabor spectrum (DGS) not only has the advantages that the discrete Wigner distribution (DWD) has, but also can efficiently eliminate the cross term defects existing in DWD. By processing a typical signal with different methods, it was found that the DGS is better than the short time Fourier transform and the discrete Wigner distribution. It has high resolution on both frequency domain and time domain without the cross term defects. With this method applied to the chinese speech processing, the results show that it can describe how the spectral content varies with time. It will open a new way for the speech recognition and the speaker identification.