In order to solve the problem of low accuracy of face recognition due to image motion blur and noise in intelligent invigilation of Electronic assessment, a face recognition model based on Gabor wavelet Convolutional neural network support vector machine is proposed. The original image is decomposed into a Covariance matrix containing amplitude and angle features by GW filtering, thus improving the performance of face recognition in fuzzy environments. An improved CNN network learning the Covariance matrix generated by GW is proposed to extract the face features. Apply SVM to classify facial feature representations and ultimately output facial recognition results. Through experimental verification, compared with PCANet, VGGFace, and ResNet50 models, the proposed GW-CNN-SVM model has better recognition performance in low resolution, motion blur, and noisy environments. The experimental results verify that the proposed GW-CNN-SVM model has higher robustness in low resolution, motion blur, and noisy environments during intelligent invigilation of school exams.