Research And Application Of Sentiment Classification Of Network Video Barrage Based On Image Analysis

Barrage is a new form of communication that has become popular in recent years and viewers can express their opinions while watching the video.However there are few research in this area at present due to the short rise of barrage,but the barrage contains a great deal of emotional information which can feedback the mood change of the audience when watching the video and is also closely related to the video itself.Emotional marking of the video keyframes according to emotions of the barrage can be convenient for the user to select the frame for video playback according to the emotion of the keyframes.The main work is completed as follows:(1)The thesis proposes a new CLSTM algorithm model for sentiment classification and establishs a five-category barrage corpus(happy,good,hate,anger,worry)by manual annotation.The CLSTM algorithm model is a new type of network structure that is based on long and short term memory network(LSTM)and convolutional neural network(CNN).In the experiment four kinds of network models were used to classify the barrage emotionally.The experiments show that our model achieves better results than the CNN,RNN and LSTM network model alone.(2)The thesis improves the hierarchical clustering algorithm and proposes a new HC-FCM algorithm for extracting keyframes.The algorithm is based on the improved hierarchical clustering and FCM clustering algorithm.In the experiment three kinds of algorithms were used to extract the key frames.Experiments show that our algorithm has better accuracy than hierarchical clustering and FCM clustering.Afterwards the key frame is emotionally marked according to the emotion of the barrage corresponding to the key frame.In the system users can choose video playback position based on keyframe sentiment.(3)The thesis designs and develops a video retrieval system based on the sentiment of the video barrage.The system includes five modules.And the algorithms introduced in the previous chapters are applied to each module.Finally the function of each module in the system has been tested.The tests show that each module in this system has good performance and normal operation which can meet the basic needs of users.Through the research on sentiment classification algorithm and key frame extraction algorithm,a new video retrieval system based on the sentiment of barrage is established.It satisfies the user's more diversified retrieval needs and realizes the value of a large amount of emotional information contained in the barrage.Users can intuitively understand the information contained in the video.