Project Description and Overview

GEORGIA INSTITUTE OF TECHNOLOGY

School of Electrical and Computer Engineering

ECE4006 Senior Design

Final Project Report

Tuesday, April 23, 2002

Digital Impulse Control Group

Wayne Blake

Mary Nsunwara

Reshun Gethers

I. Abstract

Many engineering principles are based on being able to achieve a large system gain wherein the input to the system may be small and insignificant but leads to an output that is large and complex. Implementation generally involves acquiring the signal, processing it, and executing the desired command. The input to such a system could come from the human body. Body mechanisms such as heartbeats, blood-flow, and muscle movements all have physical attributes, such as observable force and electric potential. The purpose of this project is to read in electrical signals produced by specific face muscle movements, interface these signals to a PC, and subsequently use these signals to control an end system, such as a remote control car or a wheelchair.

Signals were measured using a differential amplifier, the design of which was made available by previous design groups. Initial testing involved comparing the outputs produced by the old amplifier to the new circuit built this semester. Some undesired properties of the waveforms were observed during initial testing, including a noisy waveform during accidental movements and the presence of a high frequency component. Twisting the input wires together reduced the amount of noise and eliminated the high frequency signal.

The final leg of the project was initially supposed to be signal processing using the Keithley A/D converter. However, since this did not work as expected, the group elected to illustrate digitizing the output data using MATLAB. If the amplifier board is still working by next semester, a future group will be able to proceed with further implementation.

II. Project Description and Overview

Electromyography (EMG) is an instrumental technique for registering the electrical signals occurring when the fibers within a muscle contract on receipt of a motor command from the brain's motor cortex. In medical applications, the properties of the signals generated from the muscles can then be used to diagnose muscular or nerve dysfunctions, but this is an application that makes use of only the appearance of the signals. In certain engineering applications, however, the signals are used as inputs for electrical or mechanical systems. This is important to the procedures applied in building appropriate signal acquisition circuits.

The most simplistic description of the project assigned to group N3 this semester is to use electrical signals emitted by muscles on the head as inputs for the control of a more elaborate system, for example, a remote control car. Implementing such a system involves the following steps; reading the signals from an appropriate area of the body, amplifying and digitizing them so that the signals can be clear and useful to the end control system, transferring these clarified signals to a computer system, creating codes that direct the system to perform separate specific actions, and finally, implementing these commands on the remote control car.

For the purposes of this project, we have chosen to implement a circuit that functions much like a surface electromyogram as the measuring tool for the signals emitted by the muscles. The reasoning behind this decision is that a simple and inexpensive electromyogram can give sufficient information to power our end system, and surface, rather than needle-type contact with the skin affords us painlessness, and the freedom to place it on the body and remove it whenever necessary. Due to the superfluous level of sophistication and the cost of most the currently available electromyography systems, the decision to build our system rather than buying one was certainly prudent.

Although larger than brain signals, the signals coming from facial muscles are still relatively small, and may generally have an un-uniform analog-type waveform, and therefore are not of the appropriate form to control our end system. Thus, it is necessary to amplify these signals to a significantly larger value. Another alternative is designing the system so that the input level needed to trigger a given command is just as small as the signal coming off the face. However, since only one of these methods can be applied, a decision has to be made before design commences. The analog signals also have to be digitized so that they take on the form of impulses, which are expectedly more useful in controlling computer systems. This can be carried out by using a differential amplifier circuit, which forces the output to remain at a quiet DC level while not in use, and to give off impulses when movements occur.

The choices of facial muscles are plenteous. Signals can be acquired from the eyebrows, by blinking, or even from the muscle contractions while biting down. With surface electrodes taking signals from some or all of these muscles, there is a variety of signals being given off by the face. This equips the designers with a larger number of code choices. For instance, for the control of a motorized system, moving the left cheek could mean turn left, moving the right cheek could mean turn right, and moving both muscles could mean go straight. A choice of command assignments such as this would then leave both eyebrows available for different commands altogether. A versatile system such as this is very important for assisted daily living, seeing as most movements in everyday life are simply not in four directions only.

III. Brief Introduction to Electromyography

The contractions of all muscles are triggered by electrical impulses. These impulses can be transmitted by nerve cells, created internally by devices such as the pacemaker, or created externally with electric-shock devices. In skeletal muscles, an electric signal travels down a nerve cell, causing it to release a chemical message. This message is known as a neurotransmitter. It is released into a small gap, called the synapse, located between the nerve cell and muscle cell. The neurotransmitter crosses the gap, binds to a protein on the muscle-cell membrane and causes an action potential in the muscle cell.

Equipment for measuring the motor unit action potentials (MUAPs) associated with muscle movement are widely available. In fact, electromyographic instruments have been used for many years to detect muscle action potentials with the use of electrodes. The monitored activity can range from less than 0.1uV to as high as several thousand microvolts. For example, relaxed muscles such as those in the forehead region generally exhibit voltages in the range of 0.75 to 3 uV. Large muscles such as the quadriceps can exhibit activity as high as 2000uV. There are two types of electrodes available for performing electromyography - needle and surface electrodes. The less common of the two is the needle electrode. This electrode is utilized when specific muscle strands are to be monitored. Surface electrodes will read the activity of individual muscles or muscle groups. During this process the electrodes are placed on your skin over the muscle to be monitored. The most common muscles that are currently utilized are the frontalis located in the forehead, the masseter located in the jaw, and the trapezium.

The principal processing component in any EMG instrumentation system is the amplifier. The amplifier itself normally consists of multiple stages of amplification. However, the most important is the first stage known as pre-amplification. Together, the various stages perform numerous functions. The main function of EMG circuitry is maintaining strong signal integrity while reducing the noise power. To best reduce the noise signal, differential amplification is the method of choice for EMG circuits. The differential amplification technique is shown schematically in Figure 1. The signal is detected at two sites and the circuitry subtracts the two signals and then amplifies the difference. In other words, any signal that is "common" to both detection sites will be removed and signals that are different at the two sites will have a "differential" that will be amplified.

The accuracy with which a differential amplifier can operate is measured by the Common Mode Rejection Ratio (CMRR), which is the ratio of the differential signal gain to the common mode gain. A perfect differential circuit would have a CMRR of infinity. A CMRR of 32,000 or 90 dB is generally sufficient to suppress unwanted electrical noises. Current technology allows for a CMRR of 120 dB, but there are reasons for not trying to push the CMRR to high such as cost and stability.

Once a reasonable signal-to-noise ratio has been achieved, there are other areas of signal processing that become important depending on the task at hand. In terms of converting from low voltage signals on the surface of the skin to digital information, concepts such as filtering, sampling, and analog-to-digital conversion become significant. These concepts will be examined in much more detail once specific requirements have been identified.

Electromyography Applications Currently Available

The current use of Electromyography (EMG) is split into two main categories: biofeedback and communication. EMGs are utilized in biofeedback for clinical purposes. EMG instruments measure the electrical signal generated by the muscles. Small pads called electrodes are placed on the skin and a voltage is picked up from this electrical signal. The instrument then amplifiers the voltage and produces a tone and depending on the instrumenta bar graph, digital readout or graph on a computer screen. This sound or display can be used to teach a person to relax certain muscles; for the treatment of tension headache, sore backs and shoulders and prevention and treatment of things such as animus, high blood pressure and OOS (RSI) a painful muscle condition. Since the science of electromyography has been evolving rapidly in recent years. Another way of using the EMG in biofeedback is to strengthen muscles. The EMG is also utilized as a form of communication in controlling electronic and computer devices. However the development of such applications is relatively new, thus there are limited products that perform this task. NASA has developed an armband, which demonstrated the ability to control a 757 passenger jet simulation. The armband was implanted with eight electrodes to read muscle-nerve signals. Then the signals are sent to a computer. The computer uses software called “neral net”. This software learns the pattern of nerve signals and associates it with a particular movement.

Present Biofeedback Products

There are many biofeedback EMG devices available, however price and performance are major concerns. Most of the inexpensive do not have computer interface and can only be used for muscle training and relaxation. Thus, the EMGs with more capabilities are very expensive. Two examples are listed below.

The Muscle Monitor EMG

The muscle monitor is a tone only EMG with a single 20Hz to 600HZ bandpass and two ranges of 20uV and 200uV. It is use for learning muscle awareness and muscle relaxation training. A 9V battery powers the instrument. However this device cannot interface with a computer. This EMG is currently price at $175.00 at

MycroTrac2 EMG

The MycroTrac2 EMG is more sophisticated than the muscle monitor. The MyoTrac2 has dual channels which both provide feedback with a tone and dual LED bar graph displays. It also has a digital readout that can display the readings on both channels such as the current values or average, peak or the ratio of the channels. The instrument can be programmed to guide a person through a training program and can also store data for later printout or display on a computer. The instrument comes complete with interface and software to connect to a computer giving real time graphs of EMG and can also store the information for later recall. Powered by 3 1.5V AA size batteries the instruments flip-up case is small enough to be held in the hand. Since this EMG has so many features it is very pricey. The price of this device is $16000.00 at

Potential Applications

The most popular potential application being envisioned at this time for the completed and marketable version of our end product is the control of a wheelchair. For people who are severely incapacitated, such that they cannot even move their hands to control their manual or motorized wheelchairs, or even for people who simply feel like trying out something on cutting edge technology, moving their facial muscles, alone could be an option. This system could also be useful for applications in similar fields, such as for people who cannot move other parts of their body that would normally perform external functions. Muscle signals from able parts of the body can be used to control a system that will aid them in performing these functions.

While controlling a system such as a wheelchair might a possible implementation of our project, there has to be a less involved way to go about the day-to-day testing in the laboratory. The laboratory has a plenteous supply of sophisticated oscilloscopes, and so displaying the control waveforms on the screen is a convenient method for observing the data.

IV. Amplifier Design

The initial stages of the amplifier design involved breaking down the project into major categories. Figure 2 illustrates a composite block diagram of the implementation of the amplifier. On the front-end of this project, the main devices are the EMG amplifier, the A/D converter, and the computer software. The back-end consists of the signal to be transmitted and the performance of the remote control car. Since the group was given full access to designs from previous teams, the amount of work for each stage was considerably reduced.

Signal Acquisition

The amplifier circuit was not designed from the scratch because we currently have an EMG design in our possession. This design is based on an older version of the BrainMaster EEG monitor that was given to a previous design group by the company. A previous design team was able to get this circuit to successfully measure muscle potentials in the jaw. Thus, by duplicating their design, we expect have the same success using the circuit as an EMG. The schematic of this circuit is shown in Figure 3.

This circuit consists of two amplification stages. The input amplifier used in the first stage is the Analog Devices AD620. The circuit is designed such that this amplifier’s output should have a gain of 50. An integrator circuit is also utilized in this first stage of amplification. The integrator circuit that is being utilized as a low-pass filter and its main purpose is to provide good linearity.

The second stage of this amplifier is designed such that it has a gain of 390. Thus, the total gain of the entire circuit is 19500. The second stage also provides a frequency response from 1.7 up to 34 Hz. We considered redesigning some part of this circuit because of the current frequency response specification. Since the amplifier circuit was originally designed to detect brainwaves, a frequency response of 1.7 – 34 Hz was ideal. However, the frequency range of muscle potentials is 30 – 500 Hz. Increasing the bandwidth could be achieved by reducing the value of resistor R11 in Figure 3. However, once preliminary testing of the previous amplifier was carried out, it was determined that no adjustments were necessary. Table 1 lists some more technical attributes of the current EMG circuit we are working with.

Type: / differential
Inputs: / (+), (-), and "ground" return
Gain: / 20,000
Bandwidth / 1.7 - 34 Hz
Input Impedance: / 10 Mohms
Input Range: / 200 uV full-scale
Output Range: / 4 volts: from 0.0 to 4.0 volts
Resolution / 0.80 uV/quantum
Input Noise: / < 1.0 uV p-p
CMRR: / > 100dB

Table 1. Other technical specifications of the EMG circuit.

The circuit used as the power supply for our current prototype is shown in Figure 4. The components included a 7805 regulator, a DC power supply, and two capacitors. This circuit was designed to supply a clean and regulated signal.

Analog to Digital Conversion

An important segment of the project is the analog to digital conversion process. At the initial stages of the project, we had what we thought was a suitable PC card that implements A/D conversion in our possesion; the Keithley DAS-1701ST-DA card. Some of its key specifications are listed below in Table 2.

Number of Inputs / 4
Number of Outputs / 4
Max Sampling Rate / 160 kS/s
Input Ranges / 0 - 5V, 0 - 1V, 0 - 100mV, 0 - 20mV
Gains / 1, 5, 50, 250

Table 2. Specifications for the current A/D converter.

This particular A/D converter was used by the previous design groups. However, this component has an ISA bus interface instead of a PCI connection. Most current computers no longer have ISA interfaces, thus it was necessary for the group to acquire an A/D card that was outfitted with a PCI connection. Since the Professor was already acquinted with the A/D card used previously, it was to our advantage to order a similar model, which was equipped with the PCI connection. The Keithley KPCI-3107 A/D converter comes with DriverLINX package software. Group N1 handled installation of the Keithley card. Additionally, they ordered some parts necessary for completing the A/D conversion. Apparently, the one of the parts shipped to them was the wrong one, and so it was necessary for our group to determine some other means of implementing the amplifier commands. The new chosen method was taking the data directly from the oscilloscope, and then digitizing the instructions in MATLAB. This would be useful, because if the group is unable to complete real life implementation of the product, there will be at least a simulation that shows exactly how the output could be digitized, possibly using DSP chips.