The Construction Of A Network Audio Data Classification Label Marking And Preprocessing System

In recent years, with the rapid development of the internet and the sharp rise ofnetwork audio data, the processing requirements of multimedia data is increasinglycomplex. The acquisition and preprocessing of multimedia network data are thecommon basis of various processing application. This paper focuses on the problemsof getting the network multimedia data, stripping and decoding the audio frames, anddescription bases on the audio classification, and designing and developing a networkaudio data classification label marking and preprocessing system. The main work ofthe paper is as follows.Firstly, the job is to obtain the multimedia data from a given range of network. Ifthe network data can be downloaded directly, it’ll be downloaded and the audio partwill be stripped from it and decoded, the data will be processed as a file in thefollowing processing stage. If the network data can only be played online, the networkstreaming protocols will be analyzed and then a real-time virtual client based on thestreaming protocol (RTSP) will be generated to get the data from network. The virtualclient downloads data in the forms of streaming. The platform strip and decode theaudio frames with the help of the project named ffmpeg. Ffmpeg is used toimplement audio data stripping and decoding from the original downloadedmultimedia data.Secondly, it’s to classify and mark the decoded audio data. Support vectormachine (SVM) is used to classify the data into mute or non-mute. Then the non-mutedata is classified into speech or non-speech, and then the speech is classified into purespeech or non-pure speech; however, the non-speech is classified into music soundand the environment sound. Finally, information of classification will be appended tothe end of the audio file according to a regulation devised for description.Finally, it’s to design the general purpose processing platform system, and thetop-down module approach is used. Starting with the needs analysis, the system isdivided into seven modules: data acquisition module, program management module,decoding and file information extraction module, audio frames stripping module,audio data classification and labeling module, results output module and externalmodules mounting module. The modules proposed above work together to make theplatform a different system according to different needs.The system is developed with C++programming language. The experimentsproved that the system satisfies the expectant targets of designing.