In recent times, recognizing human emotions accurately has become crucial for enhancing human-computer interaction. Speech Emotion Recognition (SER) enables systems to interpret emotional states from speech signals, improving applications such as virtual assistants, mental health monitoring, and affective computing. However, accurately classifying emotions remains a challenge due to the complexity of speech variations. In this paper, we propose a hybrid approach that integrates traditional machine learning techniques with deep learning models to improve emotion classification. Logistic Regression (LR) and Decision Trees (DT) are used for initial feature extraction and classification, ensuring the preservation of critical speech features, while Convolutional Neural Networks (CNN)and Long Short-Term Memory (LSTM) networks are employed for deep feature learning and sequential pattern recognition. This integration allows the model to capture complex acoustic patterns and temporal dependencies, improving classification accuracy. The proposed model was trained and tested on the TESS dataset, which provides a diverse range of emotional utterances. Our integrated approach achieved impressive (98- 99 percentage) accuracy in classifying emotions, significantly outperforming traditional methods. These results demonstrate the model’s potential for improving emotion recognition, making it valuable for real-world applications in interactive AI systems and healthcare.