Surface electromyography (sEMG) signals capture electrical activity caused by muscle contractions and, therefore, have huge potential for applications in hand gesture recognition systems in prosthetics and human–computer interaction. Nevertheless, low-channel sEMG-based efficient classification of various hand gestures within a low-dimensional feature space remains challenging due to the complexity and variability of sEMG signals. In this work, the WPT is used in decomposing the sEMG signals and extracting both time-domain and frequency-domain features in an attempt to address these challenges. Moreover, this study involves the application of LDA for the purpose of selecting related features with the aim of optimizing the classification. The study applies ANN and SVM in the classification of eight different hand gestures, achieving an accuracy of 98.14% with the ANN model and 93.86% with SVM. The results indicate the effectiveness of the proposed method in improving the accuracy and reliability of the sEMG-based hand gesture recognition system.