Speech emotion recognition based on SVM and CNN using MFCC feature extraction

Zheng Zhong

doi:10.1117/12.2657244

13 January 2023 Speech emotion recognition based on SVM and CNN using MFCC feature extraction

Zheng Zhong

Proceedings Volume 12510, International Conference on Statistics, Data Science, and Computational Intelligence (CSDSCI 2022); 125101S (2023) https://doi.org/10.1117/12.2657244
Event: International Conference on Statistics, Data Science, and Computational Intelligence (CSDSCI 2022), 2022, Qingdao, China

Abstract

This research aims to accomplish speech emotion recognition in Chinese by machine learning. For the training data, this paper used CASIA Chinese emotional speech database and to train the model. For the methods of classification, the researcher compares two kinds of methods including SVM (support vector machine) and CNN (convolutional neural network). In my Data preprocessing, this researcher used the MFCC (Mei frequency cepstral coefficient) From librosa to extract the feature of the audios in CASIA database. After training the feature with these two models, the researcher obtains two accuracies that are low. To improve the accuracy, the researcher tried to change the penalty coefficient and Gamma value for SVM. For CNN, this researcher tried to add drop out layer to the CNN structure and change the L2normalizer and number of epochs to increase the accuracy. In addition, due to the limitation of SVM’s classification ability, the accuracy and the performance is very difficult to improve. Therefore, this report will mainly focus on the speech emotion recognition using CNN. After the model has trained well, the researcher tested the model using the audio recorded by me and my classmate. After improving the accuracy of the model and recorded audios with more obvious identification feature, the researcher gets a very decent result.

Citation Download Citation

Zheng Zhong "Speech emotion recognition based on SVM and CNN using MFCC feature extraction", Proc. SPIE 12510, International Conference on Statistics, Data Science, and Computational Intelligence (CSDSCI 2022), 125101S (13 January 2023); https://doi.org/10.1117/12.2657244

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
8 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Education and training

Emotion

Feature extraction

Databases

Overfitting

Machine learning

Speech recognition

Show All Keywords

Keywords/Phrases

Search In:

Publication Years