电网APP的语音文本实体多语言识别方法

Multilingual recognition method of voice and text entities in power grid APP

为了提高电网APP语音文本实体多种类语言的识别精度,提出基于语义特征融合提取的电网APP语音文本实体多种类语言精确识别技术。采用尺度分解法调节语音线谱的增强和输出参数,建立电网APP的语音参数融合模型;采用语义模糊度检测方法,实现语音识别和特征聚类;结合语义特征融合结果,建立模糊判断的语言识别模型,辨识电网APP的语音特征,精确识别电网APP语音文本实体多种类语言。实验结果表明,采用该方法识别电网APP语音文本实体多种类语言,识别精度达到90%以上,实现了精准识别,检测性能较好。

In order to improve the multilingual recognition accuracy of voice and text entity in power grid app,a multilingual exact recognition technology of voice and text entity in power grid app based on semantic feature fusion extraction is proposed.The scale decomposition method is used to adjust the enhancement and output parameters of voice line spectrum,and the voice parameter fusion model of power grid app is established.Semantic ambiguity detection method is used to realize speech recognition and feature clustering.Then,combined with the results of semantic feature fusion,a language recognition model of fuzzy judgment is established to identify the voice features of power grid app and accurately recognize multilingual voice and text entities of power grid app.The experiment results show that this method can recognize multiple languages of voice and text entities in power grid app,the recognition accuracy is more than 90%,the accurate recognition is realized,and the detection performance is good.