The Assisted Environment Information for Blind based on Video Captioning Method

Our proposed system is to help the blind to recognize the environment information. In view of the booming development of image recognition technology, many different models of recent video description methods have been developed. However, there is still a problem that the generated subtitles are not accurate enough to describe the image. In this paper, we propose a model RNN-LSTM to identify objects in a movie and generate correct subtitles. Our model uses a layer of RNN to provide the function of identifying images, and a layer of LSTM to analyze and generate subtitles. This paper uses the MSVD data set and thirty customized movie data sets for experiments, and compares the results of the S2VT model and the RNN-LSTM model on the data set.