search

ivineetm007 / drunk-detection

This repository contains the source code for the paper Dataset of Perceived Intoxicated Faces for Drunk Person Identification.
MIT License
11 stars 5 forks source link

Audio processing help #2

Closed Kushagra0409 closed 11 months ago

Kushagra0409 commented 11 months ago

Can you please provide some details on what are the parameters for audio files, for eg. sampling rate, bitrate, channels etc. that were used during training? I need to test the model on my own audio files, but I think there is an issue with the bitrate I am using. Could you help in understanding your approach to gathering the audio files?

ivineetm007 commented 11 months ago

Hi, There is no constraint on the parameters of the audio. The audio features are extracted using opensmile library which handles it out of the box. You can look at the feature extraction code in this notebook- https://github.com/ivineetm007/drunk-detection/blob/main/Audio_models/audio_DNN_feature.ipynb