Voice recognition device and method and voice interaction system and method

Abstract

The invention relates to a voice recognition device and method and a voice interaction system and method. The voice recognition device can receive voice input of a user, and can receive a first voicerecognition result output by a far-end voice recognition module after online processing of the voice input from the far-end voice recognition module; the voice recognition device also comprises a local voice recognition module which is provided with a second acoustic model constructed based on a binary neural network algorithm, wherein the local voice recognition module processes voice features extracted from the voice input at least through the second acoustic model so as to output a second voice recognition result. The voice recognition is timely and accurate, the influence of the network connection condition is small, and the user experience is good.

Description

technical field [0001] The invention belongs to the technical field of speech recognition, and relates to a speech recognition device whose local speech recognition module adopts an acoustic model constructed using a binary neural network (Binary Neural Network, BNN) algorithm, a speech recognition method, and a speech recognition device using the speech recognition device. A voice interaction system and a corresponding voice interaction method. Background technique [0002] Speech recognition devices such as vehicle-mounted speech recognition terminals require fast (i.e. real-time) and accurate (i.e. high recognition rate) recognition of the user's voice input in order to improve user experience when it is applied in, for example, a voice interaction system. [0003] The local end of the speech recognition device usually has a local recognition engine based on the acoustic model and the speech model, wherein the acoustic model is performing calculation processing on the use...

AI Technical Summary

Problems solved by technology

The acoustic model configured or used by the local recognition engine can be, for example, GMM (Gaussian Mixture Model)-HMM (Hidden Markov Model), deep neural network (Deep Neutral Network, DNN)-HMM, deep neural network ( Deep NeutralNetwork, DNN) is modeled and trained, and they all have the disadvantages of high computational overhead, poor real-time performance, and poor recognition rate.

[0004] In order to improve the user experience, speech recognition devices generally rely on a cloud recognition engine with strong computing power and accurate speech recognition (or high recognition rate) to recognize and process speech input and obtain corresponding recognition results; however, the cloud recognition engine is set in The remote end of the speech recognition device, its wireless network communication with the speech recognition device may limit the timeliness of the transmission of recognition results, etc., thereby affecting the user experience in terms of real-time speech recognition

[0005] At present, speech recognition devices generally use a hybrid recognition method of local recognition engine and cloud recognition engine to work. When the wireless network communication status of the speech recognition device is good, the recognition result of the cloud recognition engine is used, and the recognition rate is high. When the wireless network communication condition of the recognition device is poor (such as network interruption), if the speech recognition device determines that the wireless network communication condition is poor, it will switch to the local recognition engine for speech recognition to ensure real-time performance, but it is also serious Sacrifice the recognition rate

Images