论文部分内容阅读
在分析了传统语音识别方法的特点后，提出一种规则结合模型的语音识别思想，并实现了”特征空间轨迹匹配＋种经网络”方法作为例证。实验数据反映该方法能够分别规整语音信号的时域确定因素和特征域不确定因素。实践表明,该方法是一种行之有效的实现特定领域汉语语音识别的实用方法。
After analyzing the characteristics of the traditional speech recognition methods, this paper proposes a method of speech recognition based on the rules combined with the model, and realizes the method of “feature space trajectory matching + seed via network” as an example. The experimental data show that this method can respectively fix the time-domain determinant and the feature-domain uncertainty of the speech signal. Practice shows that this method is an effective method to realize Chinese speech recognition in specific fields.