Authors : Prashant Borkar, M. V. Sarode and L. G. Malik
Page Nos : 89-91
Description :
To estimate vehicular traffic density state as low, medium and heavy, cumulative vehicular acoustic signal is collected from roadside installed Omni-directional microphone followed by acoustic feature extraction using Mel Frequency Cepstral Coefficients (MFCC) for varying combination of frame size and shift size. Classification is performed using ANFC and further performance improved using feature selection (FS) where linguistic hedges are employed for FS. Consideration of multiple contiguous frames will definitely increase the accuracy but with cost of computational time and to reduce this computational time for medium to large scale datasets, SSCG is employed in this research work. In SSCG gradient estimation decreases the training time; however, it does not cause any performance degradation for the classifier.
Keywords: Traffic density, acoustic, neuro-fuzzy, scaled conjugate gradient
