Automated Dynamic Stuttered Speech Recognition System and Conversion System using Mel Filter And Enhanced Logistic Regression Model

Confidence and communication are both negatively impacted by stuttering, a speech condition that affects more than 70 million people globally. To improve fluency and intelligibility, the Stuttered Speech Recognition (SSR) system uses cutting-edge technologies such as the Weighted Mel Frequency Cepstral Coefficient algorithm and Machine Learning (ML) models to detect and fix speech defects. It can transform text into speech for assured communication and includes noise reduction for clarity. This goes beyond the scope of conventional speech recognition and holds great promise for helping stuttering people lead more fulfilling lives and be more integrated into society at large. The characteristic features of stuttering speech are an abundance of initial repetitions and prolongations. As a result, the stuttering episodes can be categorized using sound analysis. Finding specific stuttering occurrences, such as prolongations and repeats, in stuttered speech using a feature extraction technique is described in this study. To determine how effectively it can identify stuttering prolongations and repetitions, we apply the famous Mel Frequency Cepstral Coefficient (MFCC) feature extraction. Stuttering is a disease of neurodevelopment characterized by a lack of fluency in conventional speech patterns. Prolonged or repeated sounds, syllables, sentences, or phrases are examples of disfluency. Stuttering detection is challenging and complex since it involves pathology, psychology, acoustics, and signal processing, all of which make it an intriguing interdisciplinary domain study challenge. A thorough analysis of stuttering and disfluency classification methods is presented based on acoustic data as well as statistical and machine learning and deep learning approaches. This research is helpful for scholars, academicians and medical experts to identify the limitations in traditional models and to design efficient solutions for stuttered speech classification.