Research On Active Speech Signal Detection Technology In Audio Monitoring System

With the development of modern social process,more and more audio monitoring are used,for example,monitoring equipments are needed in the case of public security administrative,and the recording time may reach 24 hours or 48 hours or even longer,but the useful information is only a part of it.It will consume a lot of unnecessary human resources and time if the public security personnel listen to the recording information all the time.Suppose that there is a computer working for automatic detection,then the needed information will be extracted as soon as it is imported,which will make the work efficiency been greatly improved.Audio monitoring system also have a wide range of applications in other areas,such as in the meeting record,in the recording pen,in the tapping or frequency scanning.In this paper,the research content is to detect the speech signal and extract it,which is reserved for the purpose of listening.In this paper,we will study the active speech signal detection technology of audio monitoring system in two steps,and select the appropriate method and combine with practical to finish the final research respectively.First,the audio signal detection and extraction,which is the detection of sound and silence signal,mainly using double threshold endpoint detection technology for the extraction of voice part,the traditional double threshold method is according to the experience to set a threshold,and the results will get according to compare the value of the calculation of the parameters and a threshold,this method lacks flexibility.In this paper it was improved int the threshold setting,according to the preamble without words,so that every paragraph has its own threshold.In addition,add the postprocessing operation based on the endpoint detection to eliminate the burr phenomenon.In order to get a good auditory effection,if the two audio clips are separated by a very short time(3 seconds),the two parts of the signal areseen as one audio signal segmentation;Second,speech signal detection and extraction.Classification,to find the speech signal segments and extract them.On the basis of five category of hierarchical audio classification,extract the characteristics of the parameters and adjust the judge order to make it more suitable for audio monitoring system,and finally get the audio signal of the original voice.After the simulation and experiment of the experimental data,get the voice signals in the first step,and extract the audio signals in the second part.At the same time,generate the corresponding file in which the starting time and ending time of each segment of speech signal and the type parameters of speech are recorded so that it is easy to carry out data query.By recording and downloading various types of recording signals to validate the algorithm,the final results show that the algorithm is simple and easy to understand,with small calculation,the detection results are good.