• Kshitij Nehete
• Shriraj Gosavi
• Adesh Kadam
• Bhushan Aher
• Swati Khokale

Emotion recognition from audio sources necessitates feature extraction and classifier training. The feature vector is made up of audio signal elements that characterise speake specific properties like tone, pitch, and energy, It’s also utilised to train the classifier model to identify a specific emotion. The training and testing components of the open source dataset for North American English were manually separated. Frustration, joy, and depression were discovered using mel frequency cepstral coefficients (MFCC) to elicit speaker vocal tract data from audio recordings. These obtained feature vectors were fed into the classification model. Following the extraction of the test dataset, the classifier will identify the underlying emotion in the test audio. As training and test databases, a Northern American English acted and natural speech corpus, real-time input English speech, and regional language databases in Hindi and Marathi were used. The impacts of increasing the number of feature vectors delivered to the classifier, as well as the two methods utilised on feature vectors, are investigated. It investigates the efficiency of speech classification in Indian English, Hindi, and Marathi. For Indian English speaking, a faithfulness of 80% was achieved.

Machine Learning, Speech Emotion Recognition, SVM, MFCC, Data Analysis, SER, Feature Extraction

"Speech Emotion Recognition Using Machine Learning", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.9, Issue 5, page no.a625-a629, May-2022, Available :http://www.jetir.org/papers/JETIR2205082.pdf

"Speech Emotion Recognition Using Machine Learning", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.9, Issue 5, page no. ppa625-a629, May-2022, Available at : http://www.jetir.org/papers/JETIR2205082.pdf