IEEE Paper Template in A4 (V1)

advertisement
An FPGA-based EMG Communication Device
Xianglong Wang, Qiushu Chen, Matthew Elissa, Bharath B. Sathiyamoorthy
University of Michigan, Ann Arbor
E-mail address: micw@umich.edu
Abstract — Unfortunate situations, such as trauma, stroke, or an
aneurysm, can render a fully cognizant person unable to
communicate verbally or physically his or her wishes. In such a
scenario, it is incredibly important to have an alternative method
communication between the patient and the hospital staff and
even the patient’s family. This paper will introduce a device that
allows for EMG based binary code communication. The device
will be able to communicate on three levels. First, it employs a
physical logic-gate communication circuit. Second, it sends 8-bit
strings to a computer program that outputs simple messages.
Third, it allows the computer program to access the computer
operating system to interface with other programs.
Keywords — electromyography, EMG, FPGA, LabVIEW,
communication system, communication device
I. INTRODUCTION
People who suffer from speech-impairing conditions such
as trauma, stroke, or aneurysm may find themselves in a
hospital situation without the ability to communicate either
verbally or physically with their caretakers. This creates
drastic issues concerning power of attorney and other medical
decisions as well as smaller issues like requesting nurse aid.
However, in many cases, patients may still retain function of
specific muscle groups. The device presented in this report
takes advantage of these voluntary muscle groups and creates
a system that enables the patient to communicate by
specifically activating these muscle groups.
In order to achieve this, the first objective was to design an
interface that could isolate and condition specific EMG
signals from two separate muscle groups (right and left flexor
carpi radialis). The second objective was to analyze and
convert these analog signals into digital signals. Once this has
been achieved, our device allowed for patient communication
on three levels: a 2-bit physical communication system, an 8bit input to receive stored message outputs, and the same 8-bit
input to interface outside the program with the operating
system of the computer.
II. EXPERIMENTAL SETUP
To accomplish the objectives, we designed a four-level
hierarchy structure for organizing the system. (Fig. 1)
The first level of the system is signal acquisition and
modification. Two channels of EMG signals are captured by
the electrodes placed on muscles. The acquired signals go
through the conditioning circuit to be amplified and filtered.
The LabVIEW VI takes the amplified signal and converts it to
digital signal: “1” a valid reflex and “0” otherwise. The digital
signal is outputted from the DAQ board.
The second level of the system is hardware control. The
Fig. 1 Flow Chart of the Device
FPGA board will recognize the signal as inputted from either
of the two channels and store it as a digital “1” or “0” in an 8bit register. In one mode, two signals will be used to control
four LEDs through a decoder. The FPGA board can also
process and send an 8-bit signal to the LabVIEW to
accomplish LabVIEW and operating system control.
The third and fourth levels of the system are LabVIEW and
operating system control. The signal received by LabVIEW
will enter a case structure and find the corresponding
command to execute. The command can either be within the
LabVIEW, or through a batch file to the operating system.
A. Gathering the EMG signal
Electrodes were attached to the forearms of the subject to
gather EMG signals from both right and left arms. The
electrodes were placed to isolate the EMG activity of the
flexor carpi radialis. One electrode was placed on the flexor
carpi radialis. The other was placed on the tendons of the
brachioradialis. Just deep to the tendons were supinator
muscles in the forearm that would serve as non-activated
EMG signals for our differential amplifier. Ground electrodes
were placed on the posterior side of both wrists, on top of the
distal radioulnar joint. This electrode placement was observed
to be the most efficacious placement in order to receive clear,
specific signals. The setup is shown in Fig. 2.
B. Conditioning Circuit
The EMG signals were each fed through a conditioning circuit
which consisted of a differential amplifier, a high pass filter
and a low pass filter. The differential amplifier used was an
AD620 set with an amplification of 20. Next was the high
pass filter. The larger problem in our signal was a wandering
baseline, so the high pass filter was place immediately after
the differential amplifier. The high pass filter was a LM741
operational amplifier with a gain of 10 and a cut-off frequency
of 70Hz. Then the signal was passed through a similar
LM741 operational amplifier set to be a low pass filter with a
gain of 3 and a cut-off frequency of 150Hz. The Bode plots
for both filters can be seen in Fig. 3.
Fig. 2 Setup of EMG Signal Acquisition
DAQ board. The outputs of the FPGA board are two digital
outputs to the decoder and one digital output to the DAQ
board.
Valid digital “1” signals from the DAQ board go through
the input channels of the FPGA and become synchronized and
shortened to a length of one clock cycle. Then the signals are
recognized by different channels, pushing either a “0” or a “1”
into the 8-bit shift register. Then the FPGA works at two
different modes, switched by an on-board switch.
In physical decoder control mode, the FPGA outputs the
two least significant bits from the shift register and uses these
two signals to control a four-LED array, which indicates four
different messages. In the meanwhile, the SSD display on the
FPGA board will display “help”.
In computer control mode, the FPGA will output a
sequence of 9 bits (a “1” header following by the content in
the shift register) to the DAQ board at a 20 Hz scan rate. The
signals will be used for future signal processing. The current
content of the shift register is displayed on both the SSD
display and the LEDs on the FPGA board.
E. Decoder and LED Array
The decoder takes two bits from the FPGA board and
converts it into four different messages. The truth table of the
decoder is given by:
i1
0
0
1
1
i0
0
1
0
1
d3
0
0
0
1
d2
0
0
1
0
d1
0
1
0
0
d0
1
0
0
0
Table 1: Truth Table of Decoder
Fig. 3 Bode Plot of Band-pass Filter in Conditioning Circuit
C. LabVIEW A/D Conversion
EMG signals are fairly irregular and fluctuating. Therefore,
to identify a “valid” signal of muscle flex, the EMG signal
must be processed. The signal should also be converted into a
digital signal, since FPGA is a programmable logic circuit
board and takes digital input.
Our LabVIEW A/D Conversion VI used a 500 Hz sampling
rate. It first computed the root-mean square (RMS) value of
the signal from 25 samples prior to the current sample to
eliminate the fluctuation. Then the value was compared with a
pre-set threshold to determine the current digital output.
A 55-65 Hz band-stop filter is used when taking in the
signal to eliminate wandering noise and wire noise. (VI size
too large to fit on the report: please see the attached VI.)
D. FPGA Logic
FPGA is a programmable logic circuit board. The inputs of
the FPGA are two down-regulated digital outputs from the
Then we can map the logic equations as:
d0=i1’i0’;
d1=i1’i0;
d2=i1i0’;
d3=i1i0.
The schematics of this 2×4 decoder are given by Fig. 4 and
Fig. 5: in Fig. 4, every AND/NOT gate is replaced by the
transistor circuit in Fig.5 since there is no logic gates available.
The d0, d1, d2, d3 are directly connected to LEDs.
Fig. 4 Decoder Schematics
Please refer to the attached VI file for the block diagram.
III. RESULTS
We were able to make the whole system able to achieve
our objectives. Fig. 7 shows the final prototype of our project:
NOT
AND
Fig. 5 NOT Gate and AND Gate
F. LabVIEW Control and Operating System Control
The LabVIEW receives an 8-bit signal (originally 9 bit,
but the header 1 is useless) from the FPGA board. This is the
control signal of LabVIEW.
The first level of LabVIEW control is controlling the
output within the LabVIEW. A correlation between the 8-bit
signals between commonly used expressions was established
(see Fig. 6). Using the LabVIEW control, the patient will be
capable of performing greetings, answering yes-or-no
questions, and describing the level of their pain by numbers.
Fig. 6 Correlation between signals and messages
The second level of LabVIEW control is interfacing with
the operating system. We designed two major functions under
this context. By interfacing a command prompt in LabVIEW,
our system is capable of opening / close music player to
satisfy need for entertainment. A decryption function is also
included. LabVIEW will call a Java-based RSA encryption
system using command prompt and a batch file when the
correct private key for RSA encryption is present. The
LabVIEW will also open the file after encryption. Otherwise
LabVIEW will warn the user that they have the wrong key.
Fig. 7 Final Prototype
The following two pictures (fig. 8, fig. 9) are taken during a
test for LabVIEW control.
Fig. 8 Acquisition of an EMG Signal in LabVIEW
Chart 1 to 3: EMG signal, RMS-valued signal, Digital signal output
medical report with our system. It is noticeable that even just
using 8 bits now, our system is capable of further
sophistication and performing more complex operation.
The major sources of errors are mostly eliminated in
breadboard, in LabVIEW and during digitization. This
consists of white noise from the tissues, wire noise and
wandering noise. However, one point that is reflected in the
testing part needs our attention: we did not recognize one (out
of 20) intended reflex. For the sake of accuracy, we selected a
higher level of threshold in digitization. We have to put in
more time, to find a balance between lowering the threshold
level and increasing the potential error rate.
V. APPLICATION
Our system can be used in families or hospitals, especially
for hospitals with centralized nurse stations. With the decoder
Since the outcomes of our projects can be classified as
mode, a buzzer is potentially added to warn the nurse about
either as expected (correct output) or not as expected
urgent needs from the patient so that the nurse can arrive on
(incorrect output), we designed several tests for our system:
button-simulated FPGA program test (using button to simulate time. Our system also provides an innovative and easy way
for the patients to communicate with nurses. With an on-board
signals from the LabVIEW), LabVIEW-FPGA signal
display, there is no need for the nurse to hold a computer to let
acquisition test (use actual EMG to test A/D conversion and
the patient read what he/she has inputted. With a reset button
FPGA) and FPGA-LabVIEW signal sending/recognition test
and a send button, it becomes much easier for patients to
(send the signal from FPGA and see the corresponding output
in LabVIEW). Our design only failed to acquire a digital input double-check so that they will not make critical mistakes in
communicating. Also, FPGA-based design can be easily
one time in the LabVIEW-FPGA signal acquisition test (out
converted into IC chip design, so it is mass-production ready.
of 20). Therefore, our design is very stable and reliable.
There are a few points that our system can improve. First,
IV. DISCUSSION AND CONCLUSION
the weight of the wires might be too heavy for a patient for the
current design. Lighter, thinner wires should be used to relieve
In conclusion, our system acquires EMG signal from both
arms, recognizes them as digital signal “1” or “0” respectively, patients’ burden if available. Also, the power of our design is
stores the signal in an 8-bit binary string, and achieve different not fully explored. Eight bits can form 256 different kinds of
messages, but our system has used less than 1/8 of them. With
kinds of communication.
more built-in functions, our design will be more powerful.
With our system, three levels of EMG-based
communication are achieved. First of all, based on a logicACKNOWLEDGMENT
gate-based LED subsystem, patients can call for specific help
The authors would like to thank University of Michigan,
simply by sending a two-bit signal and lighting up LED in
Ann Arbor’s Biomedical Engineering Department for funding
nurse station corresponding to urgent help requirements.
for this project. We would also like to thank Instructor Dr. Fan
Secondly, operating on 8-bit signal, face-to-face
and Graduate Student Patrick Ingram for offering help with
communication between nurse and patient is facilitated with
this project.
our LabVIEW based communication function. Thirdly,
interacting with computer OS is also achieved. Patients are
able to open/close music player, decrypt files to check their
Fig. 9 LabVIEW Output Control
Download