Analysis and Classification of Speech of Children with Hearing Impairment in Kannada

Hearing is essential for identifying objects, communicating through sound, and supporting speech production. However, millions of people worldwide experience hearing loss, with a significant portion affecting children. Various factors contribute to hearing impairment, including genetic predispositions and environmental exposures such as infections and noise. Early detection and intervention are crucial for improving outcomes and quality of life. Treatment options range from medical interventions like antibiotics or surgery for ear infections to advanced technologies like hearing aids and cochlear implants for sensorineural hearing loss. In this work, the main aim is to develop, analyse, and classify speech data collected from normal and hearing-impaired Kannada-speaking children. STFT spectrograms are extracted and used as features that are fed to CNN classifiers to classify normal and hearing impairment speech samples. A classification accuracy of 99.5% is obtained from the presently collected database.