Detecting Human Emotions Using Machine Learning Techniques: A Comprehensive Approach

The ability to comprehend and respond to human emotions is a key component of effective interaction, yet it remains an obstacle in artificial intelligence. With speech interactions becoming increasingly common in everyday life, machines that can analyze emotions behind spoken words offer significant potential across various fields. The proposed research develops an innovative system which merges speech-to-text processing with emotion detection from transcribed speech to allow machines to interpret human emotions through spoken words. The dual-stage framework undergoes testing through various machine learning and deep learning approaches (sadness, joy, love, anger, fear, and surprise) to assess the effectiveness of various machine learning and deep learning algorithms. Moreover, it aims to simulate human-like emotional interpretation by developing a machine learning model using Natural Language Processing and Long Short-Term Memory. This approach is effective for analyzing text and speech, as it can process sequential data, helping to identify emotions expressed in spoken language. This research aims to develop solutions that improve human-computer interactions across various domains. Speech is transcribed using speech recognition tools and then for this text we use a range of machine learning and deep learning algorithms, including Naive Bayes, K-Nearest Neighbors, Logistic Regression, Support Vector Machines, Decision Trees, Random Forests, and Long Short-Term Memory, to build multiple models. These models are then tested on a separate dataset, and their performance compared. The insights gained from analyzing residents' emotional expressions could contribute to enhancing the safety of smart home systems especially for elderly or disabled people where we can get their emotions through their spoken words.