首页>
外国专利>
UTTERANCE VERIFICATION METHOD USING MULTIPLE ANTIMODEL BASED ON SUPPORT VECTOR MACHINE IN SPEECH RECOGNITION SYSTEM
UTTERANCE VERIFICATION METHOD USING MULTIPLE ANTIMODEL BASED ON SUPPORT VECTOR MACHINE IN SPEECH RECOGNITION SYSTEM
展开▼
机译:语音识别系统中基于支持向量机的多模态验证方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
This invention relates to an ignition apparatus and a verification method using the SVM model based on the multiple half speech recognition system that, in particular in carrying out the verification utterance in a voice recognition system, speech recognition module 13 of the processing result and the number of context independent phoneme model half, that is, all context independent half model with the treatment mix, context independent model adapted half, On the other hand, using a model, such as half and half model performs a VQ-based discrimination learning multiple (N) of feature parameter value is extracted, compared with after the pre-generated model to SVM training to perform the input value is greater than the acceptance threshold, and wherein the input is configured to include a speech verification module for performing a rejection is less than or equal to the threshold value, the present invention is to use more than one half model dissimilar characteristics are modeled for different input data, the reliability of each model calculated by using the SVM-based feature parameters as input, verification of fire there is a high reliability and allows you to be very effective.
展开▼
机译:本发明涉及一种使用基于多半语音识别系统的SVM模型的点火装置和验证方法,特别是在语音识别系统中执行验证话语时,语音识别模块13的处理结果和数量上下文无关的音素模型的一半,即所有具有处理组合的上下文无关的半模型,适应上下文的一半的模型,另一方面,使用模型(例如Half and Half模型)执行基于VQ的识别学习倍数( N)提取特征参数值,与预先生成的模型进行比较后,以SVM训练执行输入值大于接受阈值,并且其中输入配置为包括语音验证模块,用于执行拒绝次数较少当阈值等于或等于阈值时,本发明将使用多于一半的模型为d如果输入数据不同,则通过使用基于SVM的特征参数作为输入来计算的每个模型的可靠性,验证火灾的可靠性就很高,可以使您非常有效。
展开▼