Model based signal enhancement for impulse response measurement

• Modellbasierte Signalverbesserung für Messungen von Impulsantworten

Wang, Xun; Vorländer, Michael (Thesis advisor)

Berlin : Logos (2014)
Dissertation / PhD Thesis

In: Aachener Beiträge zur technischen Akustik 18
Page(s)/Article-Nr.: VI, 135 S. : Ill., graph. Darst.

Zugl.: Aachen, Techn. Hochsch., Diss., 2013

Abstract

Impulse response measurements that are performed outdoors are highly susceptible to uncertainties caused by the non-perfect measurement setup, the presence of background noise, and fluctuations in media such as wind and temperature drift. This work concentrates on two scenarios: the measurement of reflection coefficients of noise barriers and the influence of temperature variances in machinery cavities. Regarding the reflection coefficient measurement, an optimized microphone array is implemented to separate direct sound and reflected sound. Compared with the standard subtraction method, it is possible to obtain the reflection coefficient through only one single measurement without moving the devices between the free-field room and the sound barrier under test, and to avoid the errors resulting from an imperfect measurement setup and time variances throughout the procedure. For the purpose of de-noising, the option of using signal statistics-based source separation methods is also studied. Simulation results show that source separation can indeed reduce the background noise effect in a reverberant environment. However, it cannot exceed the performance of synchronous averaging. When the excitation signal is known and no specific knowledge about the noise can be used, averaging is the most efficient way to improve the SNR for time-invariant systems. The application of long-time averages, however, runs the risk of time variances. The possibility of phase-shift compensation in wind fluctuations is analysed here. The time-varying phase shift can be compensated for the direct sound component. The reflection coefficient measurement has more complex effects, and both the magnitude and the phase are changed. The influence of wind fluctuations in a field consisting of many reflections cannot be compensated for by using the same approach. Temperature variances also influence the accuracy of impulse response measurements. Concerning an online machine monitoring scenario, the temperature drifts during the transfer function measurement and the speed of sound varies with temperature in the machine environment. As a consequence, the impulse response is stretched along the time axis. A time-warping model is derived and applied to compensate for inter-period (slow) and intra-period (rapid) temperature variances. In this way, measurements of higher accuracy can be obtained.