Feature_Extraction_OpenSmile

Codes for extracting features of IEMOCAP with Open Smile tool kit
Ouput - csvs file or/and pickle file of extracted features

Environment

• Linux
• python 3.8

Quick Start

1. Set up OpenSmile toolkit for your server

• Download link
• OpenSmile documents : https://audeering.github.io/opensmile/

2. Clone this repo at your server
   git clone https://github.com/jade0520/Feature_Extraction_OpenSmile.git

3. Excute below
   bash run.sh

Sturcture

feature(OpenSmile) --> csv

• extract_wavs

csv --> pickle

• csvs_to_pickle_all : save all data as one pickle
• csvs_to_N_fold_pickles_labels : divede labels for cross fold validation (1/5/10 folds & independent/dependent)
• csvs_to_N_fold_pickle : save feature csvs as array pickles based on devided label folds

Reference

• ./tool/helper.py : https://github.com/KrishnaDN/speech-emotion-recognition-using-self-attention/blob/master/helper.py

Contact

📫 seo.jiyoung@dsp.inha.ac.kr
Any comments or opinions are welcome;)