Design And Implementation Of DirectShow-based Anomaly Detection System Of Digital Television

With the evolution of radio and TV industry in China continuing to accelerate, the traditional manual detection mode about digital TV error demonstrates the disadvantages, such as a time-consuming detection program, low detection accuracy and big spending detection equipment, etc, which has not been suited for the detection demands. What’s more, the error detection systems abroad are expensive. The purpose of this project is to study an automatic technical audit system sofyware, which shows the advantages of fast detection speed, high detection precision, long detection duration time and low cost in PC terminal. These can solve the disadvantages above. This article will make automated detection technology of digital TV, which almost based on computer instead of human beings, come true. The work outlined as follows:(1) Adopt DirectShow multimedia technologies to separate the TV program’s audio and video makes software’s function modular. The system designs source filter, MPEG-2 demultiplexer filter, video filter, audio filter and video renderer filter.(2) The system detects common video errors(black fields, color bar, monochrome field, static frame, YUV excess). Combining histogram analysis and error-frame image data designs the detection methods of black fields, color bar, monochrome fields, static frame and YUV excess. Besides, a method dividing image into equal size grids. We can get the grid’s gray feature to judge the error. Compare with judging the single pixel one by one, the grid method accelerates operational efficiency.(3) After analysis MPEG-2 video encoding features, propose combing differential edge detection and Canny edge detection to get the mosaic edge, then classify the macro block edge and use macro block gray features and macro block color features to judge the mosaic.(4) After understand the characteristics of Wave format, analysis and design the detection methods of silent, over loud volume, over low volume, noise, blasting volume, sound inverting). On one hand, analysising the first four error levels characteristics in the time domain judges the error. On another hand, using the Fast Fourier Transformation(FFT) to get the noise’s energy and right and left channels’ phase judges the error.Considering about the fault’s detection speed, the accuracy of the fault detection algorithms, system stability and real-time quality, the system is detected through functional test. The outcome illustrates that it’s in accordance with the demands of the practical application in the broadcasting industry. Besides, the detection accuracy is almost higher than 98% in order to effectively implement the digital TV audio and video fault detection.