Izvorni znanstveni članak

https://doi.org/10.31803/tg-20230725074848

Hand Gesture Recognition System for Hearing and Speech Impaired People Using ML

G. K. Dayananda orcid.org/0000-0001-5499-5179 ; Faculty of Electronics & Communication Engineering, Canara Engineering College, Benjanapadavu, Bantwal Taluk, Mangalore, D. K. District, 574219 Karnataka, India *
I. R. Manjunath ; Faculty of Electronics & Communication Engineering, Canara Engineering College, Benjanapadavu, Bantwal Taluk, Mangalore, D. K. District, 574219 Karnataka, India
J. G. Sreerama Samartha ; Faculty of Electronics & Communication Engineering, Canara Engineering College, Benjanapadavu, Bantwal Taluk, Mangalore, D. K. District, 574219 Karnataka, India
M. Vayusutha ; Faculty of Electronics & Communication Engineering, Canara Engineering College, Benjanapadavu, Bantwal Taluk, Mangalore, D. K. District, 574219 Karnataka, India
D. N. Pratham ; Electronics & Communication Engineering, Canara Engineering College, Benjanapadavu, Bantwal Taluk, Mangalore, D. K. District, 574219 Karnataka, India
- Sakshath ; Electronics & Communication Engineering, Canara Engineering College, Benjanapadavu, Bantwal Taluk, Mangalore, D. K. District, 574219 Karnataka, India
A. U. Shashank ; Electronics & Communication Engineering, Canara Engineering College, Benjanapadavu, Bantwal Taluk, Mangalore, D. K. District, 574219 Karnataka, India
Swasthi Prasad Shetty ; Electronics & Communication Engineering, Canara Engineering College, Benjanapadavu, Bantwal Taluk, Mangalore, D. K. District, 574219 Karnataka, India

* Dopisni autor.

Sažetak

Sign language has historically served as a primary form of communication for the deaf community. However, the lack of sign language proficiency among the general population and the limited availability of interpreters create barriers in communication between non-sign language speakers and deaf individuals. To address this issue, a proposed system aims to bridge this communication gap by assisting non-sign language speakers in recognizing and understanding American Sign Language (ASL) gestures. This system has the potential to enhance greatly communication between deaf individuals and the wider society. The process involves enhancing the quality and clarity of hand gestures by passing them through a filter. Subsequently, a classifier analyses the filtered hand gesture and predicts its corresponding meaning or category. This classification step enables non-sign language speakers to interpret accurately the intended message conveyed through these hand gestures. It is important to note that this method specifically focuses on recognizing the 26 letters of the ASL alphabet. Analysis of various datasets reveals that the enhanced network performs better than Faster R-CNN in terms of gesture categorization accuracy. Specifically, when using 175 photos for each move, the upgraded network achieves an 83% recognition accuracy, surpassing the performance of Faster R-CNN in gesture classification. Furthermore, with 600 images for each gesture, the improved network achieves an accuracy of 97. %. This finding suggests that employing a TOF camera to collect images based on depth information effectively reduces the influence from other elements during feature extraction. Overall, this system primarily focuses on recognizing fingerspelling gestures, which represent individual letters in ASL. It is commendable that efforts are being made to utilize neural networks and technology to enhance communication for individuals with disabilities.

Ključne riječi

ASL; feature extraction; gesture recognition; hand gesture; image recognition; speech and hearing impaired

Hrčak ID:

344747

URI

https://hrcak.srce.hr/344747

Datum izdavanja:

13.3.2026.

Posjeta: 437 *