Nonlinear SVM-Type Automatic Dicision Algorithm in Noisy Environment for Hammering Test System

This paper proposes the nonlinear SVM-type auto-matic decision algorithm in a noisy environment for hammering test system. The proposed method (i) suppresses the noise which is many at low frequency by using a Butterworth type high pass filter, and (ii) extraction of hammering signal from noise signal estimation based on minimum statistics method and Kalman filter algorithm with colored driving source. Furthermore, (iii) finding the optimal discriminant boundary using a nonlinear support vector machine with the first formant and first formant frequency as features obtains from cepstrum analysis results, and (iv) judged and evaluated from an automatic decision algorithm considering time change of feature. The effectiveness of the proposed method was demonstrated by computer simulation.