|
ELECO 2009 Invited Talks |
|
Seda SENAY1, Luis F. CHAPARRO1, Mingui SUN & Robert SCLABASSI1, 2, and Aydin AKAN3 1University of Pittsburgh, Department of Electrical and Computer Engineering, Pittsburgh, PA, USA 2University of Pittsburgh, Laboratory for Computational Neuroscience, Pittsburgh, PA, USA 3Istanbul University, Department of Electronics Engineering, Istanbul, Turkey Abstract – Brain-computer interfaces (BCIs) provide a way to monitor and treat neurological diseases. BCI research and development is interdisciplinary, involving neurobiology, engineering, mathematics, and computer science. An important application of BCIs is the monitoring and treatment of epilepsy, a neurological disorder characterized by recurrent unprovoked seizures, symptomatic of abnormal, excessive or synchronous neuronal activity in the brain. About 50 million people worldwide suffer from epilepsy. Electroencephalography (EEG) is a major tool for clinical diagnosis of neurological diseases and brain research. BCIs contain an array of sensors that gather and transmit data under the constrains of low-power, no clocks and minimal data transmission. Asynchronous sigma delta modulators (ASDMs) are considered an alternative to synchronous analog to digital conversion. ASDMs are non-linear feedback systems that enable time-encoding of analog signals, equivalent to non-uniform sampling. In this talk we will consider the problems of asynchronous data acquisition and transmission, and reconstruction of the original signals. An effcient reconstruction of time-encoded signals using a prolate spheroidal waveform (PSW) projection is proposed. PSWs have finite time support and maximum energy concentration within a given bandwidth. We show that by projecting non-bandlimited signals onto the space represented by orthonormal PSW basis, the required sampling rate can be reduced nearly in half. The original signal can be reconstructed from the ASDM time-encoded binary signal. We will also show how a modified orthogonal frequency division multiplexing (OFDM) technique using chirp modulation can be used to transmit an array of time encoded signals emanating from the BCI. Our method generalizes the chirp modulation of binary streams with non-uniform symbol duration. As an alternative, the ASDM can be combined with a level crossing sampler to make the sampling adapt to the signal. The ASDM provides estimates of the local mean of the signal to the level-crossing sampler which only takes a sample when the input exceeds these local mean estimates. This non-uniform sampling and reconstruction of the original signals can be connected with compressive sensing and random filtering, which emphasize signal “sparseness.” |