Characterisation of adaptive filters used in the identification process of annoying noises in vehicles

Sie sind hier: Homepage > Suche

Characterisation of adaptive filters used in the identification process of annoying noises in vehicles

Huertas, Jose Ignacio / Antelis, Javier Mauricio

To identify annoying noises (squeaks and rattles) inside vehicle cabins, it is necessary to capture the sounds present inside them and eliminate or attenuate all external noises from the obtained signal by using LMS, RLS or Kalman adaptive filters. This work aims to characterise the performance of these filters and select the one most suitable for this application. Analytical and experimental work was developed to characterise the performance of these filters in terms of filtering capacity and filtering quality. Additionally, a subjective evaluation was performed asking a group of users about the filter that best reproduces well-defined noices when they are recorded from moving vehicles. Consistently, results showed that the RLS adaptive filter, which uses a recursive least square algorithm, is the most suitable for this application. This filter offers the greatest filtering capacity and the highest filtering quality because it is a deterministic algorithm that uses all the old data weighed by a forgetting factor. In comparison with the other filters, RLS filter requires the largest amount of computational processing time. Nevertheless, this is not a limitation for the application in the process of identifying annoying noises in vehicles. Additionally, it was found that for low frequencies of the engine and of the annoying noise, all the filters exhibit the smaller filtering capacity. Under these circumstances, the filters allow the passage of the signal of interest with high contamination of related undesired signals.