TY - GEN
T1 - Contactless Body Gesture Recognition for Enhancing Non-Verbal Communication
T2 - 2nd International Conference on Microwave, Antennas and Circuits, ICMAC 2025
AU - Fatima, Aisha
AU - Hameed, Hira
AU - Saleemi, Balal
AU - Imran, Muhammad Ali
AU - Abbasi, Qammer H.
AU - Abbas, Hasan
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Deaf-mute individuals communicate through sign language, which involves hands and hand movements, body postures, and facial expressions. Despite advancements, recognizing sign language through automation continues to be a complex and emerging field of research. Current methods typically rely on sensor-based and vision-based approaches, both of which have limitations, such as privacy concerns, maintenance requirements, and sensitivity to ambient lighting conditions. Consequently, contactless sensing has emerged as a promising solution for recognizing automatic sign language. This study proposed a framework that used contactless sensing to recog-nise five specific gestures-Sad, Neutral, Fearful, Happy, and Surprised. A dataset of 150 samples is collected (each class is repeated 30 times). Afterthat, three pre-trained deep learning models: MobileNet, ResNet50, and VGGI6 are employed for the classification purpose. ResNet50 outperformed other models with 96% accuracy.
AB - Deaf-mute individuals communicate through sign language, which involves hands and hand movements, body postures, and facial expressions. Despite advancements, recognizing sign language through automation continues to be a complex and emerging field of research. Current methods typically rely on sensor-based and vision-based approaches, both of which have limitations, such as privacy concerns, maintenance requirements, and sensitivity to ambient lighting conditions. Consequently, contactless sensing has emerged as a promising solution for recognizing automatic sign language. This study proposed a framework that used contactless sensing to recog-nise five specific gestures-Sad, Neutral, Fearful, Happy, and Surprised. A dataset of 150 samples is collected (each class is repeated 30 times). Afterthat, three pre-trained deep learning models: MobileNet, ResNet50, and VGGI6 are employed for the classification purpose. ResNet50 outperformed other models with 96% accuracy.
KW - Body Gestures
KW - Contactless Sensing
KW - Deep Learning
KW - Sign Language
KW - Xethru X4M03 Radar
UR - https://www.scopus.com/pages/publications/105007437786
U2 - 10.1109/ICMAC64768.2025.11003228
DO - 10.1109/ICMAC64768.2025.11003228
M3 - Conference contribution
AN - SCOPUS:105007437786
T3 - 2025 2nd International Conference on Microwave, Antennas and Circuits, ICMAC 2025
BT - 2025 2nd International Conference on Microwave, Antennas and Circuits, ICMAC 2025
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 17 April 2025 through 18 April 2025
ER -