Addressing the issues of poor real-time performance and insufficient prediction capability in consecutive gesture recognition tasks based on surface Electromyography (sEMG),we propose an approach utilizing Gaussian Mixture Model-Hidden Markov Models (GMM-HMMs) and Viterbi backtracking.This approach leverages sliding window technique to segment the 8-channel sEMG signals,and GMM-HMMs to classify hand gestures into 4 action states:idle,ascending,steady,and descending.A refined Viterbi sliding window marginalization strategy is implemented to ensure prolonged connections between adjacent windows,enabling anticipatory prediction of subsequent gesture states.Moreover,a dynamic threshold model based on maximum likelihood is incorporated to accurately categorize gestures.In a task involving 12 consecutive two-gesture sequences completed by 8 participants,the proposed approach attained an average recognition rate of 98.1％ with a prediction time of 71 ms,significantly outperforming both the LSTM model (94.2％,309 ms) and the GRU model (93.8％,300 ms).