Ringing noise estimation and suppression by high pass filtering and wavelet packet transform

The presence of noise in the measured signal is one of the main factors tha t affects the reliability of ultrasonic testing (LIT). The ringing noise is one of the most commonly encountered types of noise in nondestructive test ing (NDT) work. Due to the multireflection of ultrasound in the probe and t he emergence as a low frequency oscillation of the entire signal, the noise affects the anomaly visibility and signal-to-noise ratio. In this work, hi gh pass filtering and wavelet packet transform are used to process ultrason ic A-scan signals with ringing noise. High pass filtering of the signals le ads to enhancement of the signal-to-noise ratio to some extent. However, be cause what the frequency filter eliminated is the low frequency component o f the entire ultrasonic signal (both the ringing noise and short echoes fro m discontinuity), the frequency filter is not an ideal method to deal with ringing noise. Wave packet transform has a superior time frequency localiza tion property It can discriminate a short echo in a long time oscillation b ackground, thus realizing the estimation of the ringing noise while preserv ing the frequency content of the discontinuity echo. The estimated ringing noise using wave packet transform provides only a pure low frequency, long term oscillation. Subtraction of the estimated noise from the original sign al provides enhancement of the signal-to-noise ratio, and the anomaly visib ility is much better than that of high pass filtering. Finally, the power s pectral density of signals with and without ringing noise is estimated. The difference between the power spectral density functions shows explicitly t hat the ringing noise mainly affects the low frequency band of the spectrum of ultrasonic A-scan signals.