laughter-detection

This library contains code and models to segment regions of laughter from an audio file. The models/ folder contains models trained on the Switchboard data set.

Please cite the following paper if you use this software for research:

Kimiko Ryokai, Elena Durán López, Noura Howell, Jon Gillick, and David Bamman (2018), "Capturing, Representing, and Interacting with Laughter," CHI 2018

Installation

git clone https://github.com/jrgillick/laughter-detection.git
cd laughter-detection
pip install -r requirements.txt

Usage

• To train a new model on Switchboard, see the main function in compute_features.py and train_model.py

• To run the laugh detector from the command line, use python segment_laughter.py <input_audio_path> <stored_model_path> <output_folder> <threshold>(optional) <min_length>(optional) <save_to_textgrid>(optional)

• e.g. python segment_laughter.py my_audio_file.wav models/model.h5 my_folder 0.8 0.1

  Parameters

  • The threshold parameter adjusts the minimum probability threshold for classifying a frame as laughter. The default is 0.5, but you can experiment with settings between 0 and 1 to see what works best for your data. Lower threshold values may give more false positives but may also recover a higher percentage of laughs from your file.

  • The min_length parameter sets the minimum length in seconds that a laugh needs to be in order to be identified. The default value is 0.2.

  • If set to True, the save_to_textgrid parameter toggles writing the identified laughter intervals to a Praat TextGrid annotation file rather than saving them in separate WAV files. The default value is False.

  Output

  • The segmenter prints out a list of time segments in seconds of the form (start, end) and saves them as a list of wav files into a folder at <output_folder>

Dependencies

• Librosa - tested with version 0.5.0
• Keras - tested with version 2.0.0