Emotion-Identification-from-raw-speech-signals-using-TDNNs

This repo contains the implementation of the paper Emotion Identification from raw speech signals using DNNs" By Mousmita Sarma, Pegah Ghahremani, Daniel Povey, Nagendra Kumar Goel,Kandarpa Kumar Sarma, Najim Dehak in Pytorch The paper is published in Interspeech 2018 Paper: https://danielpovey.com/files/2018_interspeech_emotion_id.pdf

Installation

I suggest you to install Anaconda3 in your system. First download Anancoda3 from https://docs.anaconda.com/anaconda/install/hashes/lin-3-64/

bash Anaconda2-2019.03-Linux-x86_64.sh

Clone the repo

https://github.com/KrishnaDN/x-vector-pytorch.git

Once you install anaconda3 successfully, install required packges using requirements.txt

pip iinstall -r requirements.txt

Data preperation

This steps creates manifest files for training and testing

python dataset.py --pickle_filepath  /media/newhd/IEMOCAP_dataset/data_collected_full.pickle
                 --dataset_root /media/newhd/IEMOCAP_dataset/raw_data --store_meta meta/

If you want to add your dataset, take a look at datasets.py code and modify the code accordingly

Training

This steps starts training the model.

python training_Emo_TDNN_StatPool.py --training_filepath meta/training.txt --testing_filepath meta/testing.txt
                             --input_dim 1 --num_classes 4 --batch_size 64 --use_gpu True --num_epochs 100
                             

Note that this model is based on raw waveform TDNN.

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change. For any queries contact : [email protected]

License

MIT