Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

A Novel Hybrid Deep Learning Technique for Speech Emotion Detection using Feature Engineering

Created by
  • Haebom

Author

Shahana Yasmin Chowdhury, Bithi Banik, Md Tamjidul Hoque, Shreya Banerjee

Outline

In this paper, we propose DCRF-BiLSTM, a speech emotion recognition (SER) model that recognizes various emotions (neutral, happy, sad, angry, fear, disgust, and surprise). We train the model using five datasets: RAVDESS, TESS, SAVEE, EmoDB, and Crema-D, and achieve high accuracies on individual datasets (RAVDESS 97.83%, SAVEE 97.02%, CREMA-D 95.10%, and TESS and EmoDB 100%). In particular, when all three datasets (R+T+S) are combined, the accuracy reaches 98.82%, outperforming previous studies. In addition, this is the first study to evaluate all five benchmark datasets integratedly, achieving a high overall accuracy of 93.76%, verifying the robustness and generalization performance of the DCRF-BiLSTM framework.

Takeaways, Limitations

•
Takeaways:
◦
The DCRF-BiLSTM model achieved high accuracy on various speech emotion recognition datasets, demonstrating the superiority and generalization performance of the model.
◦
By integrating and evaluating five major datasets, we overcome the limitations of previous studies and provide a more comprehensive performance evaluation.
◦
We present the potential application of the DCRF-BiLSTM model in the field of speech emotion recognition.
•
Limitations:
◦
There may be a lack of consideration for the influence of various noises or background sounds in the real environment.
◦
There may be a lack of analysis of the balance and bias of the dataset used.
◦
Additional research may be needed to determine the interpretability of the model.
👍
Made with Slashpage