Research And Implementation Of Audio Decision System Based On Signal And Content Features

Along with the development of Chinese broadcasting station, as an intermediate product, audio broadcast has been changed the whole business process of the radio broadcast. The judgment of audio program is an essential link before the program was broadcasted. It ensures the broadcast signal quality and contents safety. The audio judgment includes mute detection, level offset exceeding detection, stereo phase reverse detection, VU exceeding detection, inaudible detection, clipping detection. It can also detect the contents of the audio, such as detect the audio whether includes the sensitive information, if it includes, the system would return the time range when key words appear, so it can ensure the audio program broadcast safety.If the workers only rely on the manual way to detect the radio signal quality and contents, they would face many difficulties. The human ears can’t identify the audio contents played by the speeding. It easily caused the error detection but need a lot time and human power in the whole process.Aiming at the above problems, the theory analysis the background and understands the system requirements. Regards audio programs as the research object, I design and realize the audio decision system by using the method of the mathematical statistical, the technology of digital signal processing, speech recognition technology and so on. Compared with the traditional method, the audio decision system has some characteristics such as high efficiency, high speed and low false alarm rate.In this paper, the system would detect the quality and contents of the audio, compete three main aspects of work:First, using the FFmpeg technology to decode the audio files so that acquire the audio signal parameters such as sample rate, bit rate, number of channels, the sampling scheme and so on. Then they provide the preparation for the audio judgment.Second, it researches and realizes the audio judgment module based on the signal feature, the module mainly completes the audio quality detection, including mute detection, level offset exceeding detection, stereo phase reverse detection, VU exceeding detection, inaudible detection, clipping detection. It realizes these items detection through the provided information by Wave files or the audio sample data by transformation.Third, it researches and realizes the audio judgment module based on the contents feature, the module mainly completes the audio contents detection. Based on the i Fly mobile speech platform, it makes full recognition to the opened audio file and then searches the given key words from the recognition results, finally locates the key words.Finally, this paper makes a lot summary and gives some improvement suggestion. This audio decision system is vitrified to run well under the experiment environment, can exactly detect the substandard quality audio and location the audio key words, and reaches the previous goal.