Sign language is a form of communication that uses a visual-manual methodology of hand gestures to interact among the hearing-impaired community in meaningful conversations. The World Health Organization (WHO) estimates that more than 1 billion people need assistive technology, especially those with hearing disabilities. The advancement of digital technology is an opportunity to demonstrate the importance of real-time digital-based AT for people with hearing disabilities. By utilizing the power of deep learning and Convolutional Artificial Neural Networks (CNNs), the authors seek to develop a system capable of accurately identifying and interpreting hand signals in real time on Android mobile applications such as active prostheses for people with hearing disabilities. The author uses an unprecedented approach model that combines CNN and EfficientNetB0 architectures by utilizing ImageNet, MATLAB, and several libraries such as PyTorch, TensorFlow, NumPy, Keras, and Seaborn. The author uses a sign language detection system on Android mobile devices with Eclipse IDE for Java, JRE System Library, Android 2.2, Android Development kit, and OpenCV library and uses gesture images as binary still images. This combined method can detect strong hand gestures in American Sign Language (ASL). The prowess of the researcher's method from the real results in the study shows that the accuracy value of the proposed model is quite high because it has a success percentage value of 97.15% in training, 99.89% in testing, and 95.31% in validation. The accuracy results of this study have outperformed the accuracy value of Alexnet which is only 78%.