This study primarily focuses on integrating gait energy images (GEI) with a CNN-LSTM model, and can be roughly divided into the following four steps:

(1) Gait Feature Extraction:Firstly, the data to be tested is transformed from raw gait sequences into corresponding GEI gait silhouettes and gait energy images. The gait sequences are used for GEI transformation, accumulating the results into gait energy images representing each time segment to generate a single gait energy map.

(2) Feature Sequence Representation:Next, the GEI gait energy map sequence is used as the input data for the CNN-LSTM model. Each gait energy map is considered as the input for a time segment, forming a sequence of gait energy maps as the input data for the CNN-LSTM model.

(3) CNN-LSTM Model Training:Using the GEI gait energy map sequence as input data and corresponding data labels (e.g., gait category or detection of gait signals) as targets, the model undergoes training. The CNN-LSTM model learns the temporal correlation of gait sequences and dynamic features of the entire gait process. Parameter adjustments and training are performed on the model.

(4) Gait Analysis or Recognition:After training completion, the trained CNN-LSTM model is used for the classification or recognition of new gait sequences. The new gait sequence data is transformed into GEI gait energy maps, which are then input into the model. Gait category analysis or detection and recognition of gait signals are performed based on the model's prediction results.