Speech Features and Machine Learning Pipeline for Depression Recognition in Hikikomori Patients

This repository is part of the SOLITAIRE project. It contains a dataset package with non-identifiable speech features derived from hikikomori patients undergoing cognitive behavioral therapy (CBT), along with code developed for machine learning-based speech analysis in depression assessment. The package supports reproducible analysis of acoustic features and their association with depression-related clinical outcomes. If you use this resource, please cite both: Leal, S. S., Ntalampiras, S., Rossetti, M. G., Trabacca, A., Bellani, M., & Sassi, R. (2025). Speech-Based Depression Recognition in Hikikomori Patients Undergoing Cognitive Behavioral Therapy. Applied Sciences, 15(21), 11750. https://doi.org/10.3390/app152111750 Leal, S. S., Ntalampiras, S., & Sassi, R. (2026). Speech Features and Machine Learning Pipeline for Depression Recognition. Zenodo. DOI: 10.5281/zenodo.20099145 The resource includes: - segment-level acoustic features derived from speech recordings of 35 patients across 8 CBT sessions; - Python scripts for feature processing and machine learning-based speech analysis; - documentation describing the speech-feature columns; - citation information for the associated publication and Zenodo record. > Note: This package is associated with the broader clinical project called Solitaire. The main project DOI is 10.5281/zenodo.20526320.