A hybrid deep learning model combining convolutional neural networks with long short-term memory architecture has achieved high recognition accuracy in identifying young children by their walking patterns, according to research published in Cluster Computing. Gait recognition serves as a non-contact biometric technology designed to identify individuals from their movement dynamics over distance.
To overcome the shortage of training data specific to pediatric populations, the research team constructed a novel children’s gait dataset adhering strictly to the established CASIA-B acquisition protocol. The acquired video frames underwent preprocessing via Gaussian Mixture Modeling to binarize visual input and isolate human silhouettes.
The researchers developed an integrated neural pipeline combining spatial feature extraction with sequential temporal processing. In the initial phase, a specialized convolutional neural network extracts spatial walking characteristics from the binarized silhouettes.
Rather than relying on conventional maximum pooling layers, the convolutional network incorporates two-dimensional adaptive average pooling.
Following spatial feature mapping, the extracted representations pass sequentially into a long short-term memory recurrent network.
Extensive verification experiments were conducted across the novel pediatric gait collection as well as a combined adult and child benchmark derived from CASIA-B. Experimental results demonstrate that the hybrid system achieved a 99 percent recognition accuracy when identifying children between the ages of 3 and 5 years. Furthermore, the architecture achieved 98 percent accuracy when distinguishing walking patterns between children and adults across the combined evaluation dataset.
Beyond addressing pediatric data scarcity through the new collection, the investigation specifically tackled performance deficits seen in previous childhood gait analysis.
Subject of Research: Deep learning framework for pediatric gait recognition
Article Title: Research on gait recognition method of children based on deep learning
Article References: Du, X., Qu, Y., Li, X., Wang, Z., Li, G., Liang, W., Wang, N., & Dong, Z. (2027). Research on gait recognition method of children based on deep learning. Cluster Computing, 30(1), Article 33. https://doi.org/10.1007/s10586-026-06764-9
Image Credits: AI Generated
DOI: 10.1007/s10586-026-06764-9
Keywords: gait recognition, pediatric biometrics, deep learning, convolutional neural networks, long short-term memory, spatiotemporal dynamics, biometric identification, Gaussian Mixture Modeling, CASIA-B, computer vision, feature extraction, batch normalization, adaptive average pooling, stride dynamics, pattern recognition, neural networks
News Source: Blake Davidson. (November 20, 2026). Deep Learning Hybrid Model Accurately Identifies Young Children by Their Gait. Scienmag.



