<p>Loss of muscle power is paralysis, and the major reason is brain-injured nerve connections. Muscle electrical activity is characterized by electromyography (EMG) signals. EMG signals help classify muscular paralysis, and statistics using EMG data from various settings detect paralysis. Due to the difficulty of EMG signal processing, short research has been done on EMG-based artificial intelligence (AI) techniques for muscular paralysis classification. The AI techniques like deep learning (DL) have garnered interest in such applications. However, designing the DL-based approach for the EMG-based muscular paralysis classification is a challenging research problem. Novel DL-based automatic muscular paralysis classification (AMPC) is proposed using the efficient deep learning (EDL) model called AMPC-EDL. The AMPC-EDL model is composed of EMG signal pre-processing, feature engineering, and classification phases. In pre-processing, the input EMG signal is wavelet-based filtered to remove artifacts, noise, and missing data to improve classification. In the feature engineering phase, the one-dimensional (1D) convolutional neural network (CNN) layers are designed for the automatic feature extraction from the input pre-processed 1D EMG signal. CNN layers are designed efficiently to improve the overall classification accuracy and reduce the computational overhead, vanishing gradient, and overfitting problems. Finally, the SoftMax classifier is utilized, followed by the fully connected layer (FCL), to accurately categorize the input EMG signal as either indicative of muscular paralysis or normal. The simulation results show the AMPC-EDL model has improved the overall accuracy by 4.78% and reduced the time complexity by 17.45% compared to existing methods.</p>