A deep learning framework for gender sensitive speech emotion recognition based on MFCC feature selection and SHAP...

Area of Science:

• Artificial Intelligence
• Natural Language Processing (NLP)
• Speech Processing

Background:

• Speech communication is fundamental to human interaction.
• Machine speech processing, a subset of AI, is crucial for NLP tasks.
• Emotion recognition from speech has applications in human-computer interaction and sentiment analysis.

Purpose of the Study:

• To introduce a novel deep learning algorithm for accurate emotion recognition from speech.
• To enhance the performance of current state-of-the-art speech emotion recognition methods.
• To explore the potential of deep neural networks in classifying human emotions from vocal cues.

Main Methods:

• Utilized deep learning techniques, including Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) with Long Short-Term Memory (LSTM) units.
• Employed advanced supervised learning algorithms trained on labeled speech datasets.
• Developed a novel algorithm for classifying emotions such as happiness, sadness, anger, fear, surprise, and neutrality.

Main Results:

• Achieved up to a 15% improvement in accuracy compared to existing state-of-the-art deep learning methods for speech emotion recognition.
• Demonstrated high accuracy in understanding and predicting emotional states from speech.
• Validated the model's effectiveness on labeled datasets for various emotions.

Conclusions:

• The proposed deep learning model offers a significant advancement in speech emotion recognition.
• The system has potential for real-time applications, such as analyzing audience emotional responses.
• This research contributes to fields like media analysis, customer feedback, and human-machine interaction by enhancing speech processing capabilities.