论文部分内容阅读
文章抓住人类语音感知多模型的特点,尝试建立一个在噪音环境下的基于音频和视频复合特征的连续语音识别系统。在视频特征提取方面,引入了一种基于特征口形的提取方法。识别实验证明,这种视频特征提取方法比传统DCT、DWT方法能够带来更高的识别率;基于特征口形的音频-视频混合连续语音识别系统具有很好的抗噪性。
The article seizes the characteristics of human speech perception multi-model and attempts to establish a continuous speech recognition system based on the composite features of audio and video under noisy environment. In the aspect of video feature extraction, a feature-based extraction method is introduced. The recognition experiments show that this method of video feature extraction can bring higher recognition rate than the traditional DCT and DWT methods. The audio-video hybrid continuous speech recognition system based on the feature mouth shape has good noise immunity.