Abstract. In order to amalgamate distributing platform of traditional TV program with mobile Internet on technology and business, it is necessary to study the method of TV channel recognition based on the collaboration between mobile terminals and publishing service of TV program. Due to some advantages based on audio features, such as less processing data volume, lower complexity of signal variation, better realtime performance and non-directional sampling, some processing steps, such as data denoising, data standardization processing, sequence alignment, tolerance processing and pairwise point checking, etc., are studied for TV channel recognition based on audio features. The experimental results show that the overall performance of the proposed approach is better than those based on frequency domain and based on time domain simply because it has some advantages such as reduction of both sampling difference and sampling environment interference for various of handheld mobile terminals, reduction of transmitting interference for content distributing server-side, excellent overall performance on accuracy and efficiency, etc. The proposed approach can also be applied into data pushing, user interactive discussion, realtime vote, etc.

Abstract. In the process of generation of meta-data, speech in news programs can be translate directly into text information based on ASR technology so as to improve cataloger's efficiency and the quality of meta-data before being put in media asset management database. For the purpose of this application, four optimal strategies has been put forward to improve the accuracy of text information extraction from news programs. The experimental results indicate that this four optimal strategies are efficient in extracting text information and the developed software based on the four optimal strategies has been put into catalog for news programs.