New methods of measuring impulse responses based on carefully designed deterministicsignals can further improve the performance o ered by classical methods. In fact,these methods are particularly interesting when measuring long impulse responses asthe ones analyzed in architectural acoustics. However, the e ects of background andimpulsive noise, distortion and time-variance are known rather qualitatively. For thisreason, the ISO 18233 encourages to develop a deeper understanding of the theoreticalbases of these techniques.In this sense, this project presents an in depth analysis of two di erent methods ofmeasuring impulse responses: the linear convolution of sweep signals with the inverse lter and the circular crosscorrelation of maximum length sequences (MLS) and inverserepeated sequences (IRS).The results of this work reveal that the sweep technique can provide signi cantreduction of distortion compared to MLS/IRS technique but, unlike what is explained inthe literature, sweep signals cannot reject all distortion artifacts from the causal part ofthe impulse response. Besides, it is proved that IRS sequences are immune to distortionof even order. On the other hand, it is con rmed that synchronous averaging procedureimproves the SNR at the microphone position by 3 dB per doubling the number ofaverages. Alternatively, it is also proved that the noise contaminating the measuredimpulse response is reduced by 3 dB every time that the length of the excitation signalis doubled. In terms of impulsive noise, the sweep technique only contaminates speci cfrequency bands of the system's impulse response, whereas the MLS/IRS techniqueuniformly distributes all impulsive noise artifacts over the entire measured impulseresponse. Finally, it is also shown that MLS/IRS measurements are more vulnerableto time-varying systems than sweep measurements.

Projecte final de carrera realitzat en col.laboració amb la Technical University of Denmark