Study Of FM Radio Security System Based On Speech Recognition Technology

Broadcast security means that understand the use of the radio spectrum utilize monitoring of frequency band and channel, take advantage of certain technical means to deal with abnormal radio. Illegal radio is the most important component in abnormal broadcast. Illegal radio is a radio station which is broadcasted propaganda to society by radio frequency and set up without authorization by the relevant departments. Illegal radio has serious security risks. To investigate and deal with illegal radio is an important part of radio management work for the radio management committee.In order to find solution to the problem that finding illegal broadcasting is difficult, time-consuming and laborious. Studied the broadcast security work flow and illegal broadcast searching method, designed a method to automatically discovery illegal radio which based on speech recognition and keyword retrieval, built a hardware platform based on EM100, HE600 and an ordinary PC machine, realized automatic protection function of FM radio.1. Workflow automation. Radio monitoring is mainly cares about contents of the two areas: the use of frequency band(band scanning), and frequency point usage(IF measurement). The system realized the automatic setting of threshold line and the solution of channel occupancy in the process of frequency band scanning, to automatically discover signal in FM band. By setting the appropriate monitoring time, complete automatically alternate the frequency band scanning and the IF measurement, and the automation of the work flow of the broadcast support system is realized. Realize the automation of broadcasting system.2. Illegal radio automatic discovery. The system calls Baidu speech recognition engine, broadcast voice into text, combined with the difference between illegal radio and normal radio broadcast, designed illegal radio key Thesaurus, using the keyword search method, realize the automatic discovery of illegal radio.3. Broadcast audio segmentation and classification. During to the characteristics of long period of broadcast and audio contains invalid music for speech recognition. The segmentation and classification of audio are required before calling the recognition engine. Studied the segmentation and classification method based on endpoint detection, which is based on the length of the silent segment, the proportion of the effective segment of the voice and the proportion of the mute, the segmentation and classification of broadcast audio are realized. Complete the extraction and segmentation of the speech part of the broadcast audio, which provided conditions for speech recognition, keyword search and illegal broadcast search.When the system applied in Chengdu radio monitoring station and Xihua University, a certain number of illegal broadcast was found, which verifies the validity of the system and the method.