Speech databases for mental disorders: A systematic review

BACKGROUND: The employment of clinical databases in the study of mental disorders is essential to the diagnosis and treatment of patients with mental illness. While text corpora obtain merely limited information of content, speech corpora capture tones, emotions, rhythms and many other signals beyond content. Hence, the design and development of speech corpora for patients with mental disorders is increasingly important. AIM: This review aims to extract the existing speech corpora for mental disorders from online databases and peer-reviewed journals in order to demonstrate both achievements and challenges in this area. METHODS: The review first covers publications or resources worldwide, and then leads to the reports from China, followed by a comparison between Chinese and non-Chinese regions. RESULTS: Most of the speech databases were recorded in Europe or the United States by audio or video. Some were even supplemented by brain images and Event-Related Potential (ERP) statistics. The corpora were mostly developed for patients with neurocognitive disorders like stutter and aphasia, and mental illness like dementia, while other types of mental illness such as bipolar disorder, anxiety, depression and autism were scarce in number in database development. STRENGTHS AND LIMITATIONS: The results demonstrated that database development of neurocognitive disorders in China is much scarcer than that in some European countries, but the existing databases pave an instructive road for psychiatric problems. Also, the methods and applications of databases from the leading countries are inspiring for Chinese scholars, who are searching methods for developing a comprehensive resource for clinical studies.