首页>
外国专利>
VOICE DETECTION METHOD, PREDICTION MODEL TRAINING METHOD, APPARATUS, DEVICE, AND MEDIUM
VOICE DETECTION METHOD, PREDICTION MODEL TRAINING METHOD, APPARATUS, DEVICE, AND MEDIUM
展开▼
机译:语音检测方法,预测模型训练方法,装置,装置和媒体
展开▼
页面导航
摘要
著录项
相似文献
摘要
Provided are a voice detection method, prediction model training method, apparatus, device, and medium, belonging to the technical field of voice interaction. A multi-mode voice end-point detection method recognizes a captured face image by means of a model so as to predict whether a user has the intention to continue speaking, and, in combination with a prediction result, determines whether a collected audio signal is the end point of the voice; since on the basis of acoustic characteristics, the features of visual modality of a face image for detection are also combined, even when background noise is strong or the user pauses during speech, it is still possible to use the face image to accurately determine whether the voice signal is the end point of the voice; therefore the interference caused by background noise and pauses in speech is prevented, thereby avoiding the problem of late or premature detection of the end of the voice interaction as a result of the interference of background noise and speech pauses, improving the accuracy of detecting the end point of the voice.
展开▼