Application of signal preprocessing and blind source separation to sound power mesurement of machines

Different standards to measure the sound power of a machine can be chosen. Historically, in a place with extremely loud background noise, the intensity standard (UNE-EN ISO 9614-X) has been used because there is not enough signal-to-noise relation for the pressure method (UNE-EN ISO 374x). Nevertheless, in some cases, the interfering intensity will make the measurement impossible. To solve this problem for both techniques, a new methodology based on the use of blind source separation algorithms is proposed in order to allow an estimation of the signals of the primary sources from some mixed signals and, therefore, to calculate the sound power of the machine of interest. To make the estimated signals as similar as possible to the real signals of the machine under test, different preprocessing techniques (FFT, DCT, cepstrum or wavelets) have been used before the blind source separation algorithms. Once the tests have been conducted for every machine, it is possible to establish at least one process, or combination to obtain results similar to those obtained using the international standards. Finally, since the signals can be scaled at the output of the blind source separation algorithms, one possible signal correction scheme is analyzed.