论文部分内容阅读
语音是人类重要的生理信号,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相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目的。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量,可提高生物雷达语音信号质量。
Speech is an important physiological signal of human beings. The bio-radar can provide a new non-contact voice acquisition method by non-contact acquisition of weak information of the vibration of the laryngeal speech organs. However, there are still many kinds of noise and clutter in the echoes received by the biological radar which contain the voice information. In this paper, discrete short-time Fourier transform is used to reconstruct the signal spectrum by changing the phase spectrum while keeping the amplitude spectrum of the speech signal unchanged, so as to remove the noise components in the radar speech signal. By comparing with the classical spectral subtraction and Wiener filtering speech enhancement algorithm, this algorithm can effectively reduce the noise component under low signal-to-noise ratio and improve the biological radar signal quality.