Mixed-signal micropower VLSI systems for biomedical array signal processing

Sensor array signal processing is ubiquitous in a variety of biomedical applications. Constraints on power and size of implantable and wearable systems dictate custom design of highly integrated solutions. This dissertation presents a design methodology for low-power integrated biomedical microsystems interfacing with sensor arrays, with application to intelligent hearing aids and implantable neural interfaces.; First I present mixed-signal VLSI adaptive microsystems interfacing with miniature microphone arrays that perform acoustic source separation and localization at microwatts of power, for use in intelligent hearing aids and acoustic surveillance. Analog gradient sensing of the broadband travelling wave field at sub-wavelength scale across a planar array of sensors yields linearly mixed instantaneous observations of multiple sources, blindly separated and 3-D localized using independent component analysis. A mixed-signal VLSI implementation of gradient flow for 3-D bearing estimation has experimentally demonstrated one degree angular resolution in localizing a broadband acoustic source at 20dB signal-to-noise ratio. The chip was integrated in an Acoustic Surveillance Unit (ASU) for autonomous acoustic sensing of the battlefield with superior accuracy in localizing ground vehicles compared to commercial systems of considerably larger size and power budget. The real-time system, comprising of the gradient flow VLSI processor and reconfigurable mixed-signal VLSI implementation of independent component analysis (ICA), experimentally demonstrates perceptually clear (12dB) separation and precise localization of two speech sources presented through speakers positioned at 1.5 m from the array on a conference room table.; Second I present microsystems for implantable neural recording. Implanted neural interfaces facilitate understanding of neurological phenomena and allow for automated medical diagnostics. A 16-channel current measuring VLSI potentiostat is presented for electrochemical detection of electroactive neurotransmitters like dopamine, nitric oxide etc. The current measurement capability spans six orders of magnitude in dynamic range down to hundreds of femtoamperes through innovative micropower design in analog-to-digital conversion. Real-time multi-channel acquisition of dopamine concentration in vitro with a microfabricated sensor array demonstrates ability of the system to be integrated in implantable monitoring system.