A dog barking emotion recognition method and device based on deep learning

Abstract

The invention discloses a dog bark emotion recognition method based on deep learning, comprising the following steps: designing a cry storage layer, an endpoint detection layer, a feed-forward deep neural network model with an attention module, a convolution neural network model, and a linear integration model based on the bark of dogs; after a cry signal is stored and endpoint detection is performed on the cry signal, calculating MFCC features, training the feed-forward deep neural network model with an attention module, and judging whether the cry is a dog bark or a non dog bark; and calculating the MFCC features and Mel Filterbank features of the bark signal, training the feed-forward deep neural network model with an attention module and the convolution neural network model through cross input of the MFCC features and the Mel Filterbank features, linearly integrating four generated models, identifying ten emotions of dog bark, and visualizing the result. In addition, the invention provides a dog bark emotion recognition device based on deep learning. With the device, emotion change of dog bark can be monitored in real time, and the emotional state of dogs can be displayed remotely.

Description

technical field [0001] The invention belongs to the field of sound technology processing, in particular to a dog barking emotion recognition method and device based on deep learning. Background technique [0002] In modern family life, dogs are increasingly becoming an important part. But in the process of getting along with them, we can't understand their needs in time, such as wanting to eat, play, and go out for a walk; when they have negative emotions, such as fear, vigilance, anger, grievance, and sadness, we can't provide them in time Comfort and resolution; let alone share positive emotions (happiness) with them. In the existing technology, the vigorous development of deep learning research has greatly improved the accuracy of human speech recognition, but it is still blank in the processing of animal call signals and emotion classification. Contents of the invention [0003] In view of the problems existing in the above-mentioned prior art, the object of the pres...

AI Technical Summary

Problems solved by technology

But in the process of getting along with them, we can't understand their needs in time, such as wanting to eat, play, and go out for a walk; when they have negative emotions, such as fear, vigilance, anger, grievance, and sadness, we can't provide them in time Comfort and solve; let alone share positive emotions (happy) with them

In the existing technology, the vigorous development of deep learning research has greatly improved the accuracy of human speech recognition, but it is still blank in the processing of animal call signals and emotion classification.

Images