Low-Power Audio Input Enhancement for Portable Devices

LMS adaptive filter;Low-power processing;CADSP;Floating gate transistors

Yoo, Heejong ; Electrical and Computer Engineering

University:Georgia Institute of Technology

Department:Electrical and Computer Engineering

关键词: LMS adaptive filter;    Low-power processing;    CADSP;    Floating gate transistors;   

Others  :  https://smartech.gatech.edu/bitstream/1853/6821/1/yoo_heejong_200505_phd.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

With the development of VLSI and wireless communicationtechnology, portable devices such as personal digital assistants(PDAs), pocket PCs, and mobile phones have gained a lot ofpopularity. Many such devices incorporate a speech recognitionengine, enabling users to interact with the devices usingvoice-driven commands and text-to-speech synthesis.The power consumption of DSP microprocessors has beenconsistently decreasing by half about every 18 months, followingGene's law. The capacity of signal processing, however, is stillsignificantly constrained by the limited power budget of theseportable devices. In addition, analog-to-digital (A/D) converterscan also limit the signal processing of portable devices. Manysystems require very high-resolution and high-performance A/Dconverters, which often consume a large fraction of the limitedpower budget of portable devices.The proposed research develops a low-power audio signalenhancement system that combines programmable analog signalprocessing and traditional digital signal processing. Byutilizing analog signal processing based on floating-gatetransistor technology, the power consumption of the overallsystem as well as the complexity of the A/D converters can bereduced significantly. The system can be used as a front end ofportable devices in which enhancement of audio signal qualityplays a critical role in automatic speech recognition systems onportable devices. The proposed system performs background audionoise suppression in a continuous-time domain using analogcomputing elements and acoustic echo cancellation in adiscrete-time domain using an FPGA.