Chronic Pain Audio Dataset

The Chronic Pain Audio Dataset comprises 256 audio recordings in Spanish from patients aged between 24 and 84 years who suffer from musculoskeletal pain. The purpose of this study is to contribute a dataset for the classification of pain levels into the following categories: None (Nada), Low (Bajo), Medium (Medio) and Strong (Fuerte).
The audio recordings were collected from diverse sources to ensure representativeness and diversity. They include interviews with individuals attending medical consultations and physiotherapy treatments, as well as voluntary contributions recorded during a knee prosthesis surgery campaign.
The recordings employ a verbal scale to describe the intensity of pain experienced by patients at the time of the interview. After being recorded, the audio data underwent processing to enhance its quality, involving stages of selection, trimming, and normalisation. During these procedures, it was determined that an optimal length for usability is between 1 and 5 seconds. Volume normalisation and the removal of silences were applied to produce high-quality audio, suitable for use in training machine learning models.