Research On Audio Signal Detection Thechnology

In addition to the visual media, sound media is the most important media, accounting for about 20% of the total amount of information. In real world, the sound media exists in the form of audio stream. AAC and G.729B are the two main kinds of current audio coding systems.To remove pre-echo effects, psycho-acoustic module in audio perceptual coding model use the transient analysis to determine the transient nature of the signal to guide the adaptive switch of the long or short block. This paper analyzes the defects of the transient signals frequency-domain energy detection method based on perceptual entropy. Combined with human auditory properties and characteristics of audio coding, a simple transient analysis - time-domain energy detection method is proposed, which can quickly determine the transient nature of audio signals. Thus, a smooth signal and a transient signal use transform window with different length can be determined.A normal voice signal, in general, contains some silence, the quiet part can reach 60% ratio in the two-way conversation. In silent process, the signal transmitted into the device contains only an environmental noise. Most of the noise sources include very little information, so in a silent process, a higher compression ratio can be easily achieve. Voice activity detection technology uses digital processing technology to distinguish the sound signal in a complex noise environment into voice signal and non-voice signal. It is one of the most critical speech recognition technologies, whose performance will directly affect the correct rate of the speech recognition system. This paper describes the voice activity detection process based on the auditory characteristics, and gives a practical voice activity detection algorithm. This algorithm has a lower error rate, which is basically below 2%.