Video summarization based on face recognition and speaker verification

In this paper, we propose a video summarization system based on face recognition and speaker verification. In the proposed system, face recognition is performed first. The Adaboost approach is adopted to find out the image regions which contain human faces. We perform the Non-negative Matrix Factorization (NMF) technique to decompose the face regions into basis and corresponding coefficients. Next, we use the coefficients as features to do classification by Support Vector Machine (SVM). Simultaneously, the voice part is used to do speaker verification via GMM-SVM approach. Finally, the video summarization is processed according to the face recognition and speaker verification results. With the consideration of both sound and image parts, the proposed system shall have better performance than traditional video summarization systems.