基于短时相位谱补偿的非接触语音检测增强算法研究

语音是人类重的生理信号，生物雷达能够通过检测喉部发音器官振动的微弱信息对语音进行非接触采集，从而提供一种新型的非接触语音获取方法。但生物雷达所接收的含有语音信息的回波中还存在多种噪声和杂波。本文采用离散短时傅立叶变换，在保持语音信号振幅谱不变的条件下通过改变棚位谱来对信号谱进行重构，达到去除雷达语音信号中噪声分量的日的。通过与经典的谱减法和维纳滤波语音增强算法进行对比，此算法能够在低信噪比条件下有效地降低噪声分最，可提高生物雷达语音信号质量。

Research on Non-contact Speech Enhancement Algorithm Based on Short-time Phase Spectrum Compensation

As a kind of important physiological signals, speech signals can be detected through bioradar by detecting the weak information of human laryngeal vibration. Thus a new technology for obtaining non-contact speech signals is developed. However, various kinds of noises and clutter waves are also existed in the echoes accepted by bioradar. In this paper, discrete short-time Fourier transform was used to reduce the noise component by changing phase spectrum to reconstruct the signal spectrmn while the amplitude spectrum of speech signals remained unchanged. Compared with traditional speech enhancement algorithms such as spectral subtraction and wiener filter, the proposed algorithm can effectively reduce the noise component under the condition of low signal-to-noise ratio, which indicated that the quality speech signals detected by biological radar can be enhanced.