No abstract

No abstract

No abstract

Protection of forests and wildlife needs efficient, reliable, and real-time detection of events such as gunshots, wood cutting, distress call of animals, etc. In this paper, we propose a gunshot detection technique through acoustic signal pattern recognition utilizing Mel-Frequency Cepstrum Coefficients (MFCC). In this work, MFCC is used to extract the features of gunshots from prerecorded analog sound files. Training of the system for gunshot detection has been done using a three layer Artificial Neural Networks (ANN) using extracted parameters of acoustic signals. For the creation of the database, 150 prerecorded gunshots have been used. From the database, 80 gunshot sound samples have been used for the training of the system. Testing has been done with the remaining 70 samples in the presence of noise. The algorithm has also been tested successfully using actual gunshot in noisy environment. Efficiency of algorithm is 95 % without noise and that decreases to 85 % in the presence of noise.

No abstract