Advanced

Hardware Design of Real-Time Neural Signal Generator

Metla, Anil Kumar LU (2015) EITM01 20111
Department of Electrical and Information Technology
Abstract
Brain Machine Interface (BMI) denominates a collection of systems which interface with the Central Nervous System (CNS). Implementation of a BMI involves the detection, extraction, processing, and translation of the signals from the Central Nervous System.

A simulator to generate extracellular recordings is proposed by P.T.Thorbergsson, H. Jorntell, F.Bengtsson, M.Garwicz, J. Schouenborg, A.J Johansson in [1]. The described simulator is available as a script implemented in the numerical computing software "Matlab". To measure the performance of the BMI systems, the Matlab script is needed to be implemented on a Field Programmable Gate Array (FPGA) providing real-time signals. The first stage of the implementation on the FGPA requires a... (More)
Brain Machine Interface (BMI) denominates a collection of systems which interface with the Central Nervous System (CNS). Implementation of a BMI involves the detection, extraction, processing, and translation of the signals from the Central Nervous System.

A simulator to generate extracellular recordings is proposed by P.T.Thorbergsson, H. Jorntell, F.Bengtsson, M.Garwicz, J. Schouenborg, A.J Johansson in [1]. The described simulator is available as a script implemented in the numerical computing software "Matlab". To measure the performance of the BMI systems, the Matlab script is needed to be implemented on a Field Programmable Gate Array (FPGA) providing real-time signals. The first stage of the implementation on the FGPA requires a hardware design of the simulator. This report presents the hardware design of the Real-Time Neural Simulator.

The Real-Time Neural Simulator the original Matlab script was to be adapted
for the hardware design. The adaptation process involved replacing the Matlab closed source signal processing and mathematical functions with hardware implementable algorithms. The performances of these algorithms are successfully verified against the Matlab functions.

The Neural Simulator Matlab script is converted from floating point to fixed point implementation and the performance is verified for different word lengths. One of the quantitative measurements for comparison of the output between original script and final fixed point script was the percent of energy difference of the output signal. For a word length of 13-bits with 1 sign, 6 decimal, and 6 fractional bits, the calculated difference is less than 5 percent depicting the strong resemblance between the outputs.
Finally the hardware design of the simulator constitutes of interconnected
hardware units. At the highest level a "Data Processing Unit" generates the target and the noise recordings, "Controller" enables, disables the processing modules and routes the data between them, and finally the RAM and ROM form the memory units to store the input and processed data. (Less)
Please use this url to cite or link to this publication:
author
Metla, Anil Kumar LU
supervisor
organization
course
EITM01 20111
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Neural Simulator, FPGA, Hardware Design
report number
LU/LTH-EIT 2016-484
language
English
id
8596953
date added to LUP
2016-05-19 14:06:27
date last changed
2016-05-19 14:06:27
@misc{8596953,
  abstract     = {Brain Machine Interface (BMI) denominates a collection of systems which interface with the Central Nervous System (CNS). Implementation of a BMI involves the detection, extraction, processing, and translation of the signals from the Central Nervous System.

A simulator to generate extracellular recordings is proposed by P.T.Thorbergsson, H. Jorntell, F.Bengtsson, M.Garwicz, J. Schouenborg, A.J Johansson in [1]. The described simulator is available as a script implemented in the numerical computing software "Matlab". To measure the performance of the BMI systems, the Matlab script is needed to be implemented on a Field Programmable Gate Array (FPGA) providing real-time signals. The first stage of the implementation on the FGPA requires a hardware design of the simulator. This report presents the hardware design of the Real-Time Neural Simulator.

The Real-Time Neural Simulator the original Matlab script was to be adapted
for the hardware design. The adaptation process involved replacing the Matlab closed source signal processing and mathematical functions with hardware implementable algorithms. The performances of these algorithms are successfully verified against the Matlab functions.

The Neural Simulator Matlab script is converted from floating point to fixed point implementation and the performance is verified for different word lengths. One of the quantitative measurements for comparison of the output between original script and final fixed point script was the percent of energy difference of the output signal. For a word length of 13-bits with 1 sign, 6 decimal, and 6 fractional bits, the calculated difference is less than 5 percent depicting the strong resemblance between the outputs.
Finally the hardware design of the simulator constitutes of interconnected
hardware units. At the highest level a "Data Processing Unit" generates the target and the noise recordings, "Controller" enables, disables the processing modules and routes the data between them, and finally the RAM and ROM form the memory units to store the input and processed data.},
  author       = {Metla, Anil Kumar},
  keyword      = {Neural Simulator,FPGA,Hardware Design},
  language     = {eng},
  note         = {Student Paper},
  title        = {Hardware Design of Real-Time Neural Signal Generator},
  year         = {2015},
}