Adaptive Gain Equalizer for Speech Enhancement

Document type: Researchreports
Full text:
Author(s): Nils Westerlund, Mattias Dahl, Ingvar Claesson
Title: Adaptive Gain Equalizer for Speech Enhancement
Series: Research Report
Year: 2002
Issue: 13
ISSN: 1103-1581
Organization: Blekinge Institute of Technology
Department: Department of Telecommunications and Signal Processing (Institutionen för telekommunikation och signalbehandling)
Department of Telecommunications and Signal Processing S-372 25 Ronneby
+46 455 38 50 00
Authors e-mail: nils.westerlund@bth.se, mattias.dahl@bth.se, ingvar.claesson@bth.se
Language: English
Abstract: This report presents a noise reduction method for speech
communication where the input signal is divided into a number of
subbands that are individually weighted in time domain according
to the short time Signal-to-Noise Ratio estimate (SNR) in each
subband at every time instant. Instead of focusing on suppression
the noise, the method is focusing on speech enhancement.

The method has proven to be advantageous since it offers low
complexity, low delay and low distortion. Also, there is no need
for a Voice Activity Detector (VAD). The method is stand-alone and
works regardless of speech coding schemes and other surrounding
adaptive systems.
Summary in Swedish: Denna rapport beskriver en algoritm för bullerundertryckning i talsignaler. Algoritmen baseras på skattningar av förhållandet mellan brusnivå och talnivå i olika frekvensband i signalen. Varje frekvensband viktas enligt detta förhållande, vilket leder till ett ökat förhållande mellan nyttig talsignal och bakgrundsbrus. Metoden har visat sig vara både robust och flexibel och kräver ingen detektor av talpauser, något som är ett vanligt krav i andra bullerundetryckningsalgoritmer.
Subject: Signal Processing\Speech Enhancement
Keywords: Speech Enhancement, Adaptive Equalizer, Subband processing, Noise Reduction
URN: urn:nbn:se:bth-00186
Edit