论文部分内容阅读
当前便携式农产品市场信息采集设备缺少语音接口,且通用领域的识别算法又过于复杂,为此提出一种适用于该设备作业环境的语音识别鲁棒性方法。首先利用MMSE估计器对带噪信号进行增强,以提高输入信号的信噪比;对增强后产生的语音失真和残留噪声,再利用倒谱均值方差归一化(CMVN)方法进行补偿。实验结果表明,该联合后的算法能有效的提高系统的识别率,特别是在低信噪比(0~10 d B)环境下更为有效。
At present, there is a lack of voice interface for portable information equipment for market information acquisition of agricultural products, and general recognition algorithms are too complicated. Therefore, a speech recognition robustness method suitable for the operating environment of the equipment is proposed. Firstly, the noise signal is enhanced by using MMSE estimator to improve the signal-to-noise ratio of the input signal. The noise distortion and residual noise after enhancement are compensated by cepstral mean square error normalization (CMVN). The experimental results show that the combined algorithm can effectively improve the recognition rate of the system, especially under low SNR environment (0 ~ 10 d B).