Acoustic echo cancellation inside a truck cabin

Abstract: In the modern world, a day without telecommunication is almost unimaginable. A telecommunication system is used in almost every sector. In earlier days only luxury cars had the feature of a hands free communication but now it has become one of the basic requirements in any segment of vehicles. One problem that arises because of this type of communication is the echo scenario where the person talking can hear their own voice back on the phone. This thesis is a study about how to implement an echo cancellation system for telecommunication inside a truck. There are some earlier works about establishing an echo free communication system in a conference hall etc., however there has not been any such implementation for a truck cabin. This research is about how to study the echo characteristics of truck cabin and how an echo cancellation system can be de-signed exclusively for Volvo trucks. The project is done by making statistical assumptions and computations of the signals and understanding the properties of them. Acoustic Echo Cancellation technique is the base of this thesis, which uses the normalised least mean square (NLMS) method. To evaluate how effective the system has worked, echo return loss enhancement is used. This project studies how to set an adaptive filter that can work precisely to cancel the echo in the case of a truck cabin. Since it is with respect to a telecommunication problem, the situation of double talk where both the users talk is also taken into consideration and the method of Normalized Cross Correlation is used to handle such a situation. The implementation is tested for four truck models FM928, FM930, FH2250,FH2099 with audio recordings of different situations. The results show that the echo cancellation is done well and the expected echo return loss enhancement value of above 20 dBis obtained in majority of the conditions. With this thesis, we get a knowledge that NLMSis good for acoustic echo cancellation in a truck and how determining the right filter length can help in the process of acoustic echo cancellation.