Exploring emotion detection in Kashmiri audio reviews using the fusion model of CNN, LSTM, and RNN: gender-specific speech patterns and performance analysis

The research examines the challenge of emotion detection in Kashmiri language utilizing audio reviews. It proposes a fusion model integrating convolutional neural networks (CNN), long short-term memory (LSTM), and recurrent neural networks (RNN). The dataset is segmented into combined, male, and female subsets, and the fusion model’s efficacy is rigorously assessed. Results demonstrate promising performance, achieving an overall accuracy of 87% on the combined dataset, albeit facing difficulties in accurately categorizing certain emotions such as sadness and passive anger. Gender-specific speech characteristics significantly impact the model’s performance, with heightened accuracy in discerning emotional nuances in female speech (91%) compared to male speech (86%). The study highlights the necessity of incorporating gender-specific speech patterns into emotion detection frameworks and underscores the efficacy of fusion models in analyzing underrepresented linguistic data. These findings advance the development of emotion detection systems tailored for Kashmiri language applications and have implications for sentiment analysis and human–computer interaction domains.