Research of Gesture Recognition Based on Improved Convolutional Neural Network and Support Vector Machine

Aiming at the cumbersome feature extraction of gesture recognition, time-consuming network training, and low network recognition accuracy, an optimization algorithm based on improved convolutional neural network (CNN) and support vector machine (SVM) was proposed. The idea of multi-layer convolution in series is used to construct the neural network model, so as to obtain better ability of gesture feature extraction. Batch normalization (BN) layer is added after relu function to prevent over fitting. The fully connected layer is replaced by an adaptive global average pool (GAP) layer to reduce the network parameters. SVM classifier in machine learning is used to replace Softmax classifier, which improves the classification ability of the model. Through experimental verification, compared with the traditional CNN, the proposed model greatly improves the accuracy of gesture recognition and effectively shortens the convergence time of the model.