Abnormal action warning on encrypted-coded surveillance video for home safety

Aiming at personal privacy protection and abnormal action warning, we proposed a scheme by analyzing body ratio of a person in encrypted-coded surveillance video. The adopted coding scheme is named context intra-coding (CIC) which codes and encrypts intra-frames, simultaneously. The CIC scheme is realized based on traditional intra-coding. We can extract personal information from encrypted-coded video under the condition of personal privacy inviolateness, and then determine whether abnormal action turns up or not. Due to the limited information for human detection, we compute the body ratio of a person. In order to realize a simple and fast abnormal action warning, posture is classified and coded as three statues, namely normal status, warning status, and abnormal status. Subsequently, a sequence of posture statues will be analyzed. Experiment results demonstrate that the proposed scheme is efficient in abnormal action warning in encrypted-coded surveillance video without violating personal privacy.