Automatic Tone Removal Algorithm and DSP Implementation

Project Type: Master/Diploma Thesis
Student: Baldauf Hermann

 

 The scope of this thesis is interference cancellation, which is concerned with the removal of periodic interference signals superimposed on audio frequency signals. Two different approaches are investigated. The first one implements an adaptive filtering technique, using the normalized least-mean square algorithm (LMS) to update the cofficients of the used filter structure. Both, finite-impulse response (FIR) and infinite-impulse response (IIR) filter structures are considered. The second approach is based on examining the spectral content of the speech signal. Different spectral estimation methods are investigated, including parametric and non-parametric methods, comparing their variance and resolution. A nonparametric method, based on the iterative weighted phase averager (IWPA) algorithm is chosen to analyse the speech signal in the frequency domain. Distinction between speech and periodic interference is made using a finite state machine (FSM), removing the interference signal by conventional filtering techniques. The performance of both approaches is compared in detail. The second part of the work deals with the implementation of the proposed methods on a digital signal processor (DSP) board and compares the performance of the physical implementation with the simulation results.