<p>Verbal communication is the predominant and most essential form of interaction in daily life. Sign language, particularly Indian Sign Language (ISL) serves as the main communication method. This visual language incorporates an extensive variety of gestures and visual cues for expression. The lack of widespread knowledge about the meaning behind these gestures leads to a communication gap between the hearing-impaired communities and hearing. Consequently, there is a pressing need for an automated system to bridge this divide. Although American Sign Language has been the subject of extensive research, ISL has unfortunately not been given equal consideration. Individuals with hearing impairments rely on hand gestures to communicate. Regrettably, the vast majority of people do not understand the significance of these gestures. Therefore, a new deep model-based hand gesture recognition of ISL is proposed in this paper to perform the hand gesture recognition on ISL. The aim of the developed model is accomplished by executing the following steps such as image collection, segmentation, and edge detection and recognition. The required image for hand gesture recognition is garnered from the benchmark dataset. Thereafter, the collected images are passed to the segmentation and edge detection phase which is performed by the Adaptive thresholding based region growing and canny edge detection (ATRG-CED). Here, the threshold value and seed value of the ATRG-CEDis optimized by the Modified Tasmanian devil optimization (MTDO). The segmented and edge images along with the input image is given to the Multiscale and attention embedded residual densenet (MAERDNet) for getting the recognized outcome. The experimental analysis is carried out on the proposed deep model-based hand gesture recognition on ISL to confirm its effectiveness.</p>