Detecting Stuttering Types Using Deep Learning

Abstract

Stuttering, recognized as a prevalent speech disfluency, presents considerable challenges in achieving accurate identification. Notably, technological advancements, such as Apple’s Siri, highlight the complexities encountered in the realm of voice-controlled tools when confronted with speech disorders, including stuttering. complexities pose significant usability challenges for individuals with stuttering, impacting daily interactions and potentially leading to profound psychological implications. The overarching objective of our research is to address this stark accuracy divide by bridging the gap between stuttered and non-stuttered speech. Beyond merely enhancing the performance of automated speech recognition tools, our aim is to contribute to development of a nuanced and adaptive model that not only accurately identifies stuttering instances but also ensures seamless integration into users’ everyday lives. This endeavor holds promise not only for improving the accuracy of stutter identification machines but also for mitigating the potential psychological impact associated with the challenges they face.