论文部分内容阅读
谱减技术是语音增强的有效方法,并被成功地应用到了车辆等环境下的Robust 语音识别中。文章对将谱减技术应用到电话语音识别中的加性噪声去除上进行了尝试。实验验证该方法有利于改进电话语音识别的性能,对不特定话音84个孤立词的实验,误识率下降了21.8%。
Spectral subtraction is an effective method of speech enhancement and has been successfully applied to Robust speech recognition in vehicles and other environments. The article attempts to apply spectrum subtraction to the additive noise removal in telephone speech recognition. Experimental results show that this method is helpful to improve the performance of telephone speech recognition, and the false positive rate dropped by 21.8% for experiments of 84 isolated words with no specific speech.