论文部分内容阅读
为了给双耳听力设备佩戴者带来更好的语音可懂度,提出了一种利用双耳时间差与声级差的近场语音增强算法,该方法首先利用这两种差异来估计语音的功率谱和语音的相干函数,然后计算干扰噪声在左右耳间的头相关传输函数的比值,最后构造两个维纳滤波器。客观评价的参数显示该算法去噪效果优于对比算法而目标语音的时间差误差和声级差误差低于对比算法。主观的言语接受阈测试表明该方法能有效提高语音可懂度。结果表明,该算法在能够有效去除干扰噪声的同时,保留了目标语音的空间信息。
In order to bring better speech intelligibility to wearers of binaural hearing devices, a near-field speech enhancement algorithm using binaural time difference and sound level difference is proposed. The method first uses these two differences to estimate the speech power spectrum And the speech coherence function, and then calculate the interference noise in the head-to-ear transfer function ratio, and finally construct two Wiener filters. The parameters of objective evaluation show that the proposed algorithm is superior to the contrast algorithm in de-noising and that the temporal difference and the sound level difference of the target speech are lower than the contrast algorithm. Subjective speech threshold tests show that this method can effectively improve speech intelligibility. The results show that this algorithm can effectively remove the interference noise while preserving the spatial information of the target speech.