The B-map project, winner of a MITI call for projects
Stuttering is a fluency disorder that affects approximately 1% of adults and manifests itself through blockages, repetitions, and/or prolongations of sounds due to a loss of control of the speech muscles. This disorder is accompanied by involuntary movements such as blinking or facial tension. Its treatment relies on speech therapy, but assessing disfluencies is complex and time-consuming. Tools based on artificial intelligence attempt to automate the detection of speech accidents related to stuttering by analyzing the audio signal. However, these models face limitations, particularly due to the lack of annotated data and the high variability of stuttering between individuals. To improve accuracy, our project proposes to also use video images to detect visual cues associated with disfluencies. The combined analysis of sound and video could help refine diagnosis and help healthcare professionals better manage stuttering. Project leader: Ivana Didirkova