According to the nice paper,Speech emotion recognition using deep 1D & 2D CNN LSTM networks,the 2D CNN LSTM model was built by tensorflow2-keras modul.With training and testing in EmoDB, the model we built showed the closest conclusion comparead with the paper.
The code should run the enviroment as follow list:
name | version |
---|---|
python | 3.8 |
numpy | 1.19.2 |
tensorflow | 2.2.0 |
librosa | 0.8.0 |
scikit_learn | 0.24.1 |
Before running the code, you sholud set up the enviroment we needed by entering the following command into the terminal:
pip install -r requirement.txt
and then verify the parameter of dataset path in main.py
__EmoDB_file_path__ = 'your_dataset_path'
and finally, running!
You can dowload Berlin Database of Emotional Speech.
Any issuse should submit directly or send email msp_baiyuhe@163.com.