Compressive Sensing system for recording of ECoG signals in-vivo

AbstractThe project “Compressive Sensing system for recording of ECoG signals in-vivo” has beencarried out as challenging collaboration between the Microelectronic System Laboratoy (LSM)of the École Polytechnique Fédérale de Lausanne (EPFL, Switzerland) and the researchnanotechnology centre IMEC, (Leuven, Belgium). In this way, the context of the thesis has beenboth, academic and industrially oriented. Regarding the latest ranking from Leiden Universityhas just been released at the end of 2011, and EPFL comes in at number twelve in the worldranking and tops the table as the first non-American institution. EPFL and ETHZ, at 12th and18th place respectively, rank as the top two non-American institutions. Considering IMEC, it hasbuilt a research campus is headquartered in Leuven, but additional R&D teams in TheNetherlands, China, Taiwan, and India, and offices in Japan and the USA Belgium. It extendsover 24,400m² of office space, laboratories, training facilities, and technical support rooms,including a 300nm and a 200mm cleanroom.During the period which has taken place in EPFL, the motivation of the present project hasbeen, first of all, a deep study of the state of art of the new revealing methodology of signalcompression called Compressive Sensing. This phase has included mathematical basis toaccomplish compression, applicability scope and a wide range of different kind algorithms forthe signal recovery solution. In this way, models in Matlab and Simulink have been implementedin order to apply an efficient compression and a reliable reconstruction.In the last years, Compressive Sensing has emerged as a revolutionary compression techniquefor sparse biological signals, which are becoming a high-dense source of information inmultielectrodes arrays-based bio-systems. Due to this fact, it has been studied how applying thenovel technique of Compressive Sensing in multichannel-multipath on-chip acquisition systemfor the recording of Electrocorticography (ECoG) and Action Potentials (AP). ECoG and APneural signals have been proved to fit with the Compressive Sensing system requirements.As a conclusion of the period in LSM, the publication “Circuit-Level Implementation ofCompressed Sensing for Multi-Channel Neural Recording” has been submitted as the firstreference about what has been called Spatial Compressive Sensing (SCS), a new method tocompress signals which are distributed over an array of multielectrodes and are sparse in thespatial domain. Compressibility and reconstruction have been proved by differentimplementations in Matlab and Cadence frames. In the same publication, a new, more compactand parallel random generator system based on serial PRBSs has been submitted.During the period in IMEC, a deepest study of the circuitry integration for Compressive Sensingoperation in neural signals has been accomplished. More specifically, a novel circuitryimplementation, for the mixing and integration of the incoming signals, has been proposedaccording to an analog approach for the multipath topology.9