Montage Lz Technologies (chengdu) Co., Ltd.

Xuepeng Wang

Taibo Dong

Gang Hu

This application relates to a method for training a speech recognition model comprising: providing a speech training data set comprising a plurality of speech data items and corresponding speech tags; providing a speech recognition model to be trained comprising a convolution neural network, a first fully connected network, a recurrent neural network and a second fully connected network which are cascade coupled together, wherein each of the networks comprises one or more network layers each having a parameter matrix; and the speech recognition model processing speech data items to generate corresponding speech recognition results; and using the speech training data set to train the speech recognition model such that the parameter matrices of at least two adjacent network layers satisfies a predetermined constraint condition; and the speech recognition model trained using at least one loss function can generate speech recognition results at an accuracy satisfying a predetermined recognition target.

Method for training a speech recognition model and method for speech recognition

A beam generator, a beam generating method, and a chip are provided. The beam generator comprises a first channel, a second channel, and a signal merging module; the first channel comprises a first-channel filter, the first-channel filter is used to filter an input signal to obtain a first filtered signal; the first filtered signal comprises a desired signal; the second channel comprises: a second-channel blocking module, used to block the desired signal in the input signal to obtain a blocked signal; a compensation filter, connected to the second-channel blocking module for compensating for the blocked signal to obtain a second filtered signal; and an adaptive filter connected to the compensation filter for adaptively filtering the second filtered signal to obtain a third filtered signal; the signal merging module is for merging the first filtered signal and the third filtered signal to obtain an output signal.

Growing community of inventors

Chengdu, China

Xuepeng Wang