11.07.2015 Views

Compressive Sensing system for recording of ECoG signals in-vivo

Compressive Sensing system for recording of ECoG signals in-vivo

Compressive Sensing system for recording of ECoG signals in-vivo

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

7. Conclusions7.1. RemarksAs it is shown along the previous chapters, this project has been arranged as collaborationbetween LSM-EPFL and IMEC. Dur<strong>in</strong>g the three months <strong>in</strong> LSM, it has been carried out a deepstudy about <strong>Compressive</strong> <strong>Sens<strong>in</strong>g</strong> oriented to the implementation <strong>of</strong> a <strong>system</strong> on-chip tocompress neural <strong>signals</strong> and send it by sav<strong>in</strong>g power consumption and heat dissipation <strong>in</strong> thetransmitter.Due to the fact that <strong>Compressive</strong> <strong>Sens<strong>in</strong>g</strong> technique applied to real implementations is <strong>in</strong> anearly phase, an extensive state <strong>of</strong> the art has been accomplished <strong>in</strong> order to be able to specifythe known limitations <strong>of</strong> the designs based on <strong>Compressive</strong> <strong>Sens<strong>in</strong>g</strong>. This study encompassesall the actual trends <strong>of</strong> design, both <strong>in</strong> digital and analog doma<strong>in</strong>, and so it is considered a goodstart<strong>in</strong>g po<strong>in</strong>t <strong>for</strong> future related projects. It has been scheduled the circuit implementation <strong>of</strong> ananalog-liked multi-path/multi-channel compression <strong>system</strong>, <strong>in</strong> order to deepen <strong>in</strong>to the feasibility<strong>of</strong> this configuration <strong>in</strong>stead <strong>of</strong> the digital implementation, which is the most used one evenwhen the available power consumption comparison between digital and analog approacheshave revealed not to be enough conclusive to dismiss the latter one.For this purpose, <strong>for</strong>emost, a complete multichannel implementation has been developed both<strong>in</strong> Matlab and Simul<strong>in</strong>k. In order to carry through this task, a neural signal, (AP) has beenconsidered as sparse <strong>in</strong>put. The reconstruction has been accomplished by apply<strong>in</strong>g BPDN andLASSO reconstruction methods, <strong>for</strong> which the mathematical bases <strong>of</strong> <strong>Compressive</strong> <strong>Sens<strong>in</strong>g</strong>have been studied.In order to implement a parallel and more compact random generator block to implement themeasurements matrix, a new design has been settled up <strong>in</strong> Cadence 5, by us<strong>in</strong>g the technologyUMC 0.18μm, <strong>for</strong> a variable number <strong>of</strong> multiple outputs.Under the scope <strong>of</strong> the regular operation <strong>of</strong> <strong>Compressive</strong> <strong>Sens<strong>in</strong>g</strong>, a new approach to thismethod has been developed by realiz<strong>in</strong>g that sparsity conditions are fulfilled <strong>in</strong> an array <strong>of</strong>multielectrodes. That it, if all the sensors are simultaneously sampled, accord<strong>in</strong>g to the usualspikes propagation among neurons, just few <strong>of</strong> these electrodes receive a peak, what results <strong>in</strong>a spatial sparse signal which can be compressed by <strong>Compressive</strong> <strong>Sens<strong>in</strong>g</strong>. This approach hasbeen implemented <strong>in</strong> Matlab and Cadence, and results have been submitted <strong>in</strong> the paper [31].Dur<strong>in</strong>g the period <strong>in</strong> IMEC, <strong>in</strong> order to implement the mixer and the <strong>in</strong>tegrator blocks which areneed <strong>in</strong> each <strong>of</strong> the paths (rows) <strong>of</strong> a channel, a comb<strong>in</strong>ed design has been implemented <strong>in</strong>Cadence 6, by us<strong>in</strong>g the technology TSMC018.In order to achieve a feasible analog implementation, different <strong>in</strong>tegrators have been tested.Integrators <strong>in</strong> cont<strong>in</strong>ues time doma<strong>in</strong> have been proved not to be an optimal solution due to the71

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!