Speech emotion recognition method for extracting depth space attention characteristics based on spectrogram

Abstract

The invention discloses a speech emotion recognition method for extracting depth space attention characteristics based on a spectrogram, which comprises the following steps : a, preprocessing speech in a standard emotion database marked with specified emotion labels to generate a spectrogram; b, sending the speech spectrogram into an AItti model to extract SEF characteristics; c, sending the speech spectrogram into a speech emotion model to extract shallow features; d, taking the SEF feature and the shallow feature as input, and sending the input to a CSWNet to generate a calibration weight feature; and e, sending the calibration weight characteristics to a network layer behind the CRNN, extracting depth emotion characteristics, and performing emotion classification through a Softmax classifier to generate a final classification result. Compared with the traditional model, the method disclosed by the invention has the advantages that the average recognition rate is improved by 8.43% onthe premise of slightly increasing the complexity of the model, the non-similar emotion distinguishing effect is obvious, and the generalization property is good.

Description

technical field [0001] The invention belongs to the field of artificial intelligence emotion recognition, in particular to a speech emotion recognition method for extracting deep space attention features based on spectrograms. Background technique [0002] With the rapid development of artificial intelligence technology and the robot industry, people continue to put forward higher requirements for the interaction of artificial intelligence. However, most artificial intelligence systems so far cannot recognize and respond to various human emotions. At present, the research on human-computer interaction mainly focuses on image emotion recognition and speech emotion recognition (SpeechEmotion Recognition, SER), while image emotion recognition is mainly aimed at the recognition of human facial expressions. Speed ​​and other issues, which lead to technically accurate implementation is limited, and image emotion recognition has high hardware requirements and is difficult to be wi...

AI Technical Summary

Problems solved by technology

[0004] In the study of the emotional features of speech emotion recognition, many scholars have made many attempts, such as the selection of traditional features, the selection of specified features to simulate the recognition effect; the selection of processed features (such as first-order difference, etc.) to simulate the recognition effect, but these sequences There are certain limitations in features, that is, taking such frequency-domain features pays attention to the frequency dimension and ignores the influence of the time dimension. On the contrary, selecting time-domain features will ignore the influence of the frequency dimension, and at the same time makes the potential properties hidden between the time-frequency domain ( associativity, etc.) are ignored

Images