A wireless system for multi-channel transmission of EEG Signals

A miniaturized wireless data acquisition and real-time signal analysis system formonitoring and analysis multi-channel EEG signal is presented.. The system includes: multi-channel EEG pr

A wireless system for multi-channel transmission of EEG Signals

by

Bin YU

A thesis submitted in partial fulfillment of

the requirement for the degree of

This thesis was approved by

Professor Nader Behdad

Assistant Professor, Electrical and Computer Engineering

University of Wisconsin – Madison

Signature : Date : a

Name : Nader Behdad

A miniaturized wireless data acquisition and real-time signal analysis system formonitoring and analysis multi-channel EEG signal is presented The system includes: multi-channel EEG probes, second order 60Hz band-stop filters, instrumental differential amplifierswith 100dB CMRR, 12bit resolution analog to digital converters, low power consumptionmicrocontrollers, Zigbee wireless point to point communication modules, serial portcommunication module, graphic user interface software and digital signal processing toolbox.Key performance includes: long range communication (50 meters), miniaturized transmitter unit(40mm*40mm*10mm), high ADC sampling rate (400 samples/channel/second) and low powerconsumption of the transmitter unit in working mode (40mW) Spectrum measurement of thefilter and 8 channel real-time wireless data transmission test are applied on this system Byimplement the miniaturized transmitter unit on the EEG probes, the cumbersome cable could beremoved Applications related in medical research and commercial products could be exploredbased on this wireless brain signal acquisition system, such as brain controlled games anddisease diagnose

First and foremost, I would like to offer my sincerest gratitude to my advisor, ProfessorNader Behdad, who has supported me with his patience and wisdom for my entire graduatestudent life My Master thesis would not have been possible without his inspiration and effort

I would like to thank Professor Justin Williams and Professor Mark Allie for theirsupport, giving me access for the equipments in their laboratory These equipments greatlyaccelerated my research process

Special thanks to Tom Richner for providing me guidance and suggestions for thisproject

I would like to thank my lab mates: Mudar A l-Joumayly, Meng Li, Eric Meunier,Dhananjaya Rao, and Hang Yang for their assistance and useful suggestions

Especially, I would like to thank my parents for their support, encouragement and love

Table of Contents

Abstract ……… i Acknowledgements ……… ii Table of Contents ……… iii

2.4 Analog to digital converter

2.5 Point to point wireless communication

Chapter3 LabVIEW Graphic user interface Design

3.1 Serial Port communication between PC and microprocessor 3.2 IIR digital filter design

3.3 File I/O between LabVIEW and MATLAB

Chapter 4 Conclusion

Chapter 5 References

Chapter 6 Appendixes

Chapter 1 Introduction

1.1 Overview

Electroencephalographic (EEG) signal records the electrical activity of the neurons nearthe scalp within the brain [1] Both physiological and pathological information could be obtainedfrom EEG signal The study of EEG signal has its applications in diagnose and treatment of braindiseases, neuroscience, and cognitive science [2-3] Such as: psychogenic non-epileptic seizures,syncope, sub-cortical movement disorders, migraine variants, catatonia, adjunct test of braindeath, prognosticate in patients with coma

EEG recording method could be categorized into two groups: invasive electrode andnoninvasive electrode Mathematical model of the brain is built up, and localization of theelectrodes has been carefully studied to build up a 3-D EEG map of the brain [4-5], in order tocover the important spots on the scalp

The invasive EEG test is far more complicated than noninvasive EEG First of all the headmust be kept absolutely still when the electrodes are plugged into the brain, because anymovement will damage the brain neuron by the electrodes The patient’s head will be placed in aframe which is pinned to his/her skull, in order to reduce the movement of the head After this,the neurosurgeon will drill several holes on the skull with great precision, plug in the electrodesand read out the data

A deficiency of the invasive EEG acquisition method is it usually took more than onemonth for the patient to recover completely from the surgery [6-7] The advantage of thisinvasive method is its high accuracy and sensitivity The signal to noise ratio of invasive EEG is

from 10 to 100 times higher than non-invasive EEG recording method Currently, invasive EEGsignal recording method emphasis in brain disease diagnoses The doctor will carefully evaluatethe necessity to apply invasive EEG record, according to the non-invasive EEG measurementrecord of the patient.

The invasive EEG method is only applied when the doctors cannot diagnose the diseaseusing the noninvasive method In other word, the noninvasive method has more application andmore acceptable by the patient Noninvasive EEG recording method does not need the doctorplug the electrodes into the patient’s brain The EEG signal could be collected by placing theelectrodes on the surface of scalp [8-10] Sometimes the conductive gel is used to improve thesignal to noise ratio (SNR)

The EEG recording and analysis system is a complex data acquisition system, whichusually includes probes, analog front end, wireless transmission and data analysis system

First of all the non invasive electrodes were placed on the scalp to collect brain signal.After the electrode would be the analog front end module, which has different layers ofamplifiers and filters, to reduce the noise and amplify the EEG signal [11-12] The amplifiersystem usually consists of the following components Buffer amplifiers used to transfer theinput/output impedance, differential pre amplifiers aimed to eliminate the common mode and60Hz noise, instrumental amplifiers used to provide a high gain to amplify the EEG signal

The filter system usually consists of DC block filter/high pass filter, 60Hz band stop filter,and low pass filter to reduce high frequency noise, includes the higher order of harmonic of

After reducing the noise and amplifying the brain signal to appropriate amplitude, ananalog to digital converter is used to convert the analog signal to digital signal Subsequently acompact low power radio frequency transceiver is used to transmit the digitalized EEG signal[14-15]

Practical EEG signal acquisition usually involves in large amount of data For example, atypical EEG data acquisition system has 32 channels, 256 samples/second/channel, and thesample period of practical experiment usually last for hours EEG data compression becomesnecessary not only for reducing the storage space, but also for shorten the sample periodic andincreasing the sample ratio Data compression technology is usually applied to large amount ofdata during wireless data transmission [16-18] The most common way to compress data is thewavelet transform, the advantages of this algorithm include: high compression efficiency andlow computational cost

An EEG monitoring and analysis system will receive the transmitted the signal via serialport connected on the desktop computer [19-20] All these hardware components will worktogether in the data acquisition system, to ensure the EEG signal could be accurately sampled,and transmitted to remote computer

Advanced signal analysis system is built in the remote computer to perform complexanalysis Such as spectrum analysis, principle component analysis, signal segmentation,independent component analysis, chaos and dynamic analysis, filtering and averaging, eventrelated potentials, and pattern classification [21-24]

Feed back and control system is also a key component in EEG signal acquisition andanalysis system One category is EEG – based control of reaching to visual targets For example,

BCI2000 system could allow people input characters using brain signal without any additionalmotor actions [25-26] Another category is EEG-based control of mechanical objects Forexample, the neuron scientists classifies EEG signal into different categories, such as move left,right, forward and backward, to control an toy car or an robot hand [27-30] Control andfeedback system focus on the practical applications of analyzing EEG signal, and helps peopleregain part of the lost body movement functions.

In this thesis, a wireless data acquisition and real-time signal analysis system formonitoring multi-channel EEG signal is presented This system consists of: non-invasive EEGelectrodes, DC filters, 60Hz band stop filters, differential amplifiers, instrumentation amplifiers,analog to digital converters, Bluetooth wireless point to point communication modules, serialport communication modules (for those laptops without a serial port, an serial to USB portconverter is provided), LabVIEW based graphic user interface and MATLAB based signalanalysis

This system could provide a wireless bridge connecting electrodes on the scalp and EEGanalysis algorithms in the computer The prototype for both 8 channels EEG and 16 channelsEEG is designed and fabricated The performances of this proposed system are as follows.Communication range: 10m, maximum wireless data transmission rate: 250k bps, number ofEEG channels: 8/16 channels, data acquisition rate: 400 sample/channel/second, common moderejection ratio: 100dB, 60Hz noise suppression ratio: 50dB For 8 channels application, the size

of the transmitter is 40*40*10 mm, and for 16 channel applications, the size is 80*100*20 mm.The power consumption is smaller than 40mw for both 8 and 16 channels applications

1.2 Organizations of the chapters

A general overview of the recent research about EEG recording and analysis was presented inChapter 1 Chapter 2 gives an introduction on the EEG signal analog front end and transmitterdesign Filters, amplifiers, and wireless transceiver are introduced in three sections in Chapter 2.Chapter 3 presents the introduction of graphic user interface and signal processing system based

on LabVIEW Finally, Chapter 4 presents a summary of this wireless data acquisition and time signal analysis system Beyond this, applications and future work is also included in thischapter

real-Chapter 2 Analog front end for EEG signal

the wireless module in charge of point to point communication between EEG front end and PC.

Figure 1 Analog front end system schematic.

Figure 2 The proposed filters and amplifiers System.

Figure 3 Analog to digital converter integrated in the Microcontroller.

2.2 Miniaturized 60Hz hum noise filter design

60Hz Hum noise comes from the outside environment of the data acquisition system Such

as coupling from the power lines, connectors, and especially form the fluorescent tube light bulbs

in the lab For example, a one meter length wire, in a lab with 9 ordinary fluorescent tube lightbulbs, will bring in 100 mV hum noise at 60Hz, which is sufficient enough to bury the EEGsignal

Current wired EEG acquisition system usually use 24-bit high resolution analog to digitalconverters to sample the signal transmitted from the cable directly and then eliminate the 60Hzcomponent using a digital filter However, a 24-bit resolution analog to digital converter willextra cost on power, sampling time, and wireless communication bandwidth A simpler andeasier way to solve this noise problem is to build a 60Hz band stop analog filter, to eliminate the60Hz noise before the analog to digital converter Then a 12-bit analog to digital converter issufficient enough to sample the filtered signal, which is twice faster than a 24-bit analog todigital converter

Passive analog filters are usually used to reduce the noise Since they are passive filters,they doesn’t bring in noise from the power source, and especially suitable for handing smallsignal In order to design a 60Hz band stop passive filter, large resistance, capacitance, andinductance value are required to lower the frequency to 60Hz Since fabricate an inductor withbig inductance value, small equivalent resistance, and small size is very difficult, a filter withonly resistor and capacitor seems to be the only practical and economical choice

Figure 4 schematic of the first order Twin-T filter

satisfy the ratio in Fig.4 Theoretically, high gain/rejection ratio can be obtained, due to the lowimpedance at resonant frequency Practically, due to the practical variations between differentresistors and capacitors, there is about -30dB rejection ratio at 60Hz for the filter we build

In order to reduce the 60Hz hum noise even more, a second order Twin-T filter is designedand fabricated The second order Twin-T filter is composed of two first order Twin-T filter inseries arrangement These first-order Twin-T filters have different center frequencies: 55Hz and65Hz As it is shown in Figure 3, the attenuation ratio at 60Hz could reach 50dB In addition, the-50dB attenuation bandwidth could reach 20Hz However, as a tradeoff the high attenuation at60Hz, this second-order filter is it has a relatively high attenuation from 10Hz to 30Hz, whichcontains useful EEG information Time domain response is also included in figure 1: the input isthe green line, while the output is blue line It could be observed from the time domain responsethat: the low frequency components have been attenuated a lot, and the amplitude of the outputsignal becomes 10 times smaller, compared to the input signal The second-order filter is onlyused in complicated electromagnetic environment

Figure 5 Measurement Result of the first-order Twin-T filter

Figure 6 Measurement Result of the second-order Twin-T filter

2.3 Instrumental Amplifiers

The EEG signal is around 20 micro-volts for non invasive electrodes, buried in the 60Hznoise signal, which is around 20 milli-volts Then a differential instrumental amplifier becomesthe best choice to amplify the EEG signal Usually, 60 to 100dB of voltage gain (1000 to 10,000times) is chosen to amplify the EEG signal to a suitable range for the analog to digital converter

In addition, a common mode rejection ratio is required up to 100dB, to suppress the 60Hz noisecomponents Preferred requirements also include low power consumption, small size, and lowoffset voltage

The differential amplifier ISL28470 is a feasible solution for all these requirements It has 4differential channels, and gain range from 10 to 10000, with 110 dB common rejection ratio,within a standard TSSOP28 package Figure 7 is the external circuit connections of ISL 28470

2.4 Analog to digital Converters

Currently the highest ADC resolution rate in the market is 24bit The raw signal is sampledwithout an analog filter, and then a digital filter is applied on the sampled data to eliminate the60Hz component precisely Usually the 60Hz noise is the fundamental component of raw EEGsignal, the signal to noise ratio is around 1/300

The presented analog filter has -50dB attenuation rate at 60Hz, which means that the 60Hzcomponent will be eliminated by this filter, and signal to noise ratio could reach 1/0.3 after theanalog filter In another word, a 24bit ADC has the same resolution rate as a 16bit ADC with60Hz band stop analog filter before it

Since the 60Hz noise signal is reduced, a 10bit analog to digital converter with analog filter

in its front, is sufficient to sample the EEG signal (since 8 bit is the minimum requirement) Inthis project, 12 bit analog to digital converter is chosen as a demo The analog to digitalconverter resolution could be easily adjusted form 8bit to 14bit, by modifying a parameter in theprogram

It is interesting to mention that, there is a tradeoff between analog to digital converterresolution rate and sampling rate Namely, it takes more time to finish a sample periodic forhigher analog to digital converter resolution:

Single ADC conversion time= (clock period) * (resolution bits being converted)

For 8 channel EEG applications, Chipcon’s product CC2430, with 8 channels integratedanalog to digital converter, is chosen to realize our EEG system For 16 channel EEG