Method and device for extracting speech features based on artificial intelligence
AssigneeBaidu Online Network Technology (Beijing) Co., Ltd.
InventorsChao Li, Xiangang Li
Life of this patent
Estimating the term…
Abstract
Embodiments of the present disclosure provide a method and a device for extracting speech features based on artificial intelligence. The method includes performing a spectrum analysis on a speech to be recognized to obtain a spectrum program of the speech to be recognized; and extracting features of the spectrum program by using a gated convolution neural network to obtain the speech features of the speech to be recognized. As the spectrum program can describe the speech to be recognized in a form of image, and the gated convolution neural network is an effective method for processing images, the speech features extracted with this method may accurately describe characteristics of the speech.
Claims20
A method for extracting speech features based on artificial intelligence, wherein the speech features are used for speech recognition, and the method is performed by at least one computing device and comprises:
- performing, by the at least one computing device, a spectrum analysis on a speech to be recognized, to obtain a spectrum program of the speech to be recognized;
- and
- extracting, by the at least one computing device, features of the spectrum program by using a gated convolution neural network to obtain speech features of the speech to be recognized;
- wherein the gated convolution neural network comprises a plurality of convolutional layers, a pooling layer is connected at behind of each of the plurality of convolutional layers;
- each of the plurality of convolutional layers comprises at least two layers, an output of a former of the at least two layers is an input of an adjacent posterior of the at least two layers, each of the at least two layers comprises a first channel and a second channel, and the first channel and the second channel employ different non-linear activation functions.
A device for extracting speech features based on artificial intelligence, comprising:
- a memory, a processor and computer programs stored in the memory and executable by the processor, wherein the processor is configured to execute the computer programs to perform acts of:
- performing a spectrum analysis on a speech to be recognized, to obtain a spectrum program of the speech to be recognized;
- and
- extracting features of the spectrum program by using a gated convolution neural network to obtain speech features of the speech to be recognized;
- wherein the gated convolution neural network comprises a plurality of convolutional layers, a pooling layer is connected at behind of each of the plurality of convolutional layers;
- each of the plurality of convolutional layers comprises at least two layers, an output of a former of the at least two layers is an input of an adjacent posterior of the at least two layers, each of the at least two layers comprises a first channel and a second channel, and the first channel and the second channel employ different non-linear activation functions.
A non-transitory computer readable storage medium, with computer programs stored, wherein the computer programs are executed by a processor to achieve a method for extracting speech features based on artificial intelligence, the speech features are used for speech recognition, and the method comprises:
- performing a spectrum analysis on a speech to be recognized, to obtain a spectrum program of the speech to be recognized;
- and
- extracting features of the spectrum program by using a gated convolution neural network to obtain speech features of the speech to be recognized;
- wherein the gated convolution neural network comprises a plurality of convolutional layers, a pooling layer is connected at behind of each of the plurality of convolutional layers;
- each of the plurality of convolutional layers comprises at least two layers, an output of a former of the at least two layers is an input of an adjacent posterior of the at least two layers, each of the at least two layers comprises a first channel and a second channel, and the first channel and the second channel employ different non-linear activation functions.
Dependencies are inferred from the claim text.