Automatic Emotion Recognition System: A cross Culture Study between Tamil and Russian Speaking Children

Speech based communication contains both linguistic and paralinguistic information. It conveys emotions not only through its verbal communication, but also through voice characteristics such as pitch, volume, and stress. The objective of this study is to convey the child’s emotion to a caretaker. To achieve this, a cross-cultural study of four emotions "joy - neutral (calm state) - sadness - anger" in children aged 8-12 years is considered. The data collection is carried out in multiple modes like acting speech in Russian and spontaneous and acting speech in Tamil language. SVM and MLP classitiers were used by the Russian specialist to train and validate speech data of both languages. The accuracy achieved by the Russian team is 84.1% and the Indian specialists used CNN and obtained the training and validation accuracy of 90% and >68% over 75 Epoch and used 7-layer CNN and obtained the training and validation accuracy of 95% and > 65% over 50 Epoch. The goal of the study is to see how well Russian and Indian specialists can recognise the emotional condition of Russian and Indian children based on their speech, and the results are correlated with computer-based emotion classification algorithms with various machine learning and deep net architectures.