Audio Classification Event Identification Based On SVM Technology

With the help of audio information recognition technology,the bad information and harmful information mixed in the query of massive audio information on the network can be automatically detected by audio recognition,which not only greatly reduces the labor cost but also is more effective.Identify harmful information more quickly.Therefore,for today's network society,the application level of audio recognition technology is very extensive.Audio recognition technology is not only suitable for the Internet,it also has a wide range of applications in real life.With the development of artificial intelligence,researchers have shown great interest in audio scene understanding,and audio scene classification and event detection have also become the focus.Audio scene event recognition is a specific application based on audio recognition.The purpose is to determine the occurrence of certain events based on specific audio.Compared with traditional video surveillance and human monitoring,in some specific cases,it can make up for the shortcomings of the above methods.Compared with the traditional video surveillance technology,the screaming sound recognition device developed by Maoxia is not affected by the environment such as night and fog,and can greatly reduce the workload of the monitoring personnel,and there is no problem of the video blind zone.At present,the main monitoring systems are mostly video surveillance systems.The video surveillance system is extremely susceptible to the light environment and it is difficult to play its due role at night.The timeliness of monitoring depends largely on the monitoring personnel.In the case of high fatigue or inattention of monitoring personnel,the dangerous details in video surveillance are easily missed.Most of the major monitoring systems are video surveillance systems.The video surveillance system is extremely susceptible to the light environment and it is difficult to play its due role at night.The timeliness of monitoring depends largely on the monitoring personnel.In the case of high fatigue or inattention of monitoring personnel,the dangerous details in video surveillance are easily missed.In addition,there is an inevitable blind spot problem in video surveillance.A traditional video surveillance system cannot capture any information in the blind area of the line of sight.Audio monitoring can effectively overcome various innate problems in video surveillance and has high robustness.In this experiment,the method of audio recognition is applied to the characteristics of forests in the scene of preventing forest piracy,and the audio information is collected in real time through the point coordinate distribution of the sound sensor.The application of audio recognition to the forest can avoid the consultation of the trees.The information collected by the distributed audio sensor can determine the specific coordinates of the event according to the location of the collector.And the audio monitoring is full-time and uninterrupted,which is safer and more accurate than the existing human-powered mountain.The experiment uses the Mel spectral cloistral coefficient as an audio feature and extracts it after audio prepossessing.The support vector machine is used as a classifier to perform training recognition on a total of 200 audio data including five types of chainsaw sound,hand saw sound,engine sound,machine roar,and wind noise.The final overall recognition rate is 99.1%,and the recognition rate of the five types of audio data has reached more than 90%,which proves the feasibility of applying audio recognition technology to the identification of forests.