Voice emotion recognition system and method based on machine learning

Abstract

The invention discloses a voice emotion recognition system and method based on machine learning. The voice emotion recognition system comprises a recording noise reduction module, a punctuating module, a speaker recognition module, a feature extraction module and an emotion recognition module. The punctuating module is used for receiving recording data transmitted by the recording noise reductionmodule and cutting the recording data into fragments according to relevant features of the phonetics. The speaker recognition module is used for receiving the fragments transmitted by the punctuatingmodule, classifying the fragments through a machine learning algorithm and recognizing a speaker according to classification. The feature extraction module is used for receiving the fragments transmitted by the punctuating module, extracting the frequency spectrum features and the Mel-frequency cepstral coefficients from the fragments, and extracting the fragment features after processing the frequency spectrum features and the Mel-frequency cepstral coefficients. The emotion recognition module is used for receiving the fragment features generated by the feature extraction module, training emotion prediction models through the machine learning algorithm, and integrating prediction results of all the models through an integration algorithm. The voice emotion recognition system and method have the beneficial effect of being effectively represented in the Chinese language environment and the actual production environment of customer service telephones well.

Description

technical field [0001] The invention relates to the technical field of speech recognition, in particular to a machine learning-based speech emotion recognition system and method. Background technique [0002] In the field of speech recognition, it can be roughly divided into two modules, one is based on converting the content expressed by speech audio into text form, and the other is based on speech audio, recognizing the emotions contained in the audio (for example: anger or calm, etc.). Regarding speech emotion recognition, it has been involved in foreign literature, but the limitations are relatively large; in the existing literature, most of them are emotion recognition in languages ​​other than Chinese, and cannot be directly applied to speech in a Chinese environment to recognize emotions , and most of the literature is to use a single algorithm to recognize emotions. The recognition effect is relatively closer to the laboratory data, and the performance in actual prod...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the above-mentioned technical problems in related technologies, the present invention proposes a machine learning-based speech emotion recognition system and method

Images