Detecting adhd from speech using full-band and sub-band convolution fusion network

ADHD is a neurodevelopmental disorder marked by changes in social development and communication, however, the lack of medical psychiatrists worldwide often causes delays in diagnosing ADHD. To tackle this problem, recent progress in artificial intelligence has facilitated the use of data such as fMRI and EEG in ADHD detection. Nevertheless, the acquisition of such data can be prohibitively expensive. This study is based on the collected audio signals in collaboration with an NHS Foundation Trust in the UK. We fuse the information from the full-band and sub-band audio signals to aid in the detection of ADHD. Encouragingly, the proposed method demonstrated promising performance with an average accuracy of 87.66 %.