Change search
ReferencesLink to record
Permanent link

Direct link
Crosstalk cancellation in a FM radio system using blind source separation methods
2004 (English)Independent thesis Advanced level (professional degree), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

This thesis deals with applications of Independent Component Analysis (ICA) and Blind Source Separation (BSS) methods for a crosstalk cancellation in a FM radio transmission system. The FM radio is used for broadcasting speech based newspapers to visually impaired persons. Crosstalk occurs when two audio channels (left and right), in a FM-transmission, leaks over to each other and causing a small but undesirable mixture of both audio channels. To be able to develop algorithms to separate the mixed signals, linear identification methods are used to identify the mixing system. Based on the identification results a modified on-line Maximum Likelihood (ML) algorithm is derived to separate the undesirable mixture. The ML approach is chosen because of its simplicity and its un-need for pre-whitening. The suggested algorithm is derived for both instantaneous- and convolutive mixtures. When deriving the algorithm, different design criteria are included to fit a real-time fixed point environment better. The suggested algorithm based on instantaneous mixtures is implemented on a 16-bit Digital Signal Processor (DSP), which also serves as a platform for the newspaper receiver. The real-time implementation performs very well for both real- and generated signals. Also the implemented algorithm is made adaptive to cope with an eventually time dependent environment. Performance simulations for both convolutive and instantaneous mixtures are presented. For the instantaneous mixture case the performance is also measured under different noise- and mixing conditions. For the convolutive case a separation of 40-50 dB is measured after 90 seconds of speech. For the instantaneous mixture case a separation of 70-80 dB is measured after 20 seconds of speech, in a well conditioned high SNR environment. Using our adaptive real-time DSP implementation the separation, for the same algorithm using generated signals, is about 50-60 dB after 10 seconds of speech. For real signals the separation performance is based on human listening tests. For these signals, using our real-time implementation on the DSP, the crosstalk effect is almost impossible to notice with a pair of decent speakers.

Place, publisher, year, edition, pages
Keyword [en]
Technology, blind source separation, bss, independent component analysis, ica, crosstalk, dsp, digital signal processor, real-time
Keyword [sv]
URN: urn:nbn:se:ltu:diva-46274ISRN: LTU-EX--04/030--SELocal ID: 3e9748e7-29bf-40cd-99b8-f8f68e2993f7OAI: diva2:1019587
Subject / course
Student thesis, at least 30 credits
Educational program
Computer Science and Engineering, master's level
Validerat; 20101217 (root)Available from: 2016-10-04 Created: 2016-10-04Bibliographically approved

Open Access in DiVA

No full text

Search outside of DiVA

GoogleGoogle Scholar

Total: 6 hits
ReferencesLink to record
Permanent link

Direct link