Model of Circuits Used to Operate Motors

The signals sent from the DAQ are collected by 2 different circuits that each operate a motor. These circuits each include a SPST (single pole, single switch) relay and a DPDT (double pole, double throw) relay. The output signal from the DAQ triggers the transistor on each relay and switch the relay. Please see the figure below for more detail. For every relay coil seen on the right has connections as seen on the left in figure 20.

Figure 19: Diagram showing motors and circuitry

Motion Artifact During Surface EMG/EOG Collection

There are several sources of motion artifact that can interrupt the EOG signal that is being acquired. The main ones that must be filtered out or taken account of are those created by neck movements (see figure 17), blinking of the eyes (see figure 16), or EMG signals from surrounding muscles of the eyes. Another cause of motion artifacts can be the involuntary movement of the electrodes on the subject’s face as well as the movement of the subject himself. In order to account for these artifacts in the signal reading, their frequency range must be filtered out. The frequency range of these motion artifacts are small however, not that much smaller than that of the EOG signal. Therefore to stop the crossing over of the two frequencies the artifacts must all be filtered out. Below are images of the eye blinking and neck movement signals that will be filtered out using the MATLAB software [9].

Figure 16: Effect of eye blinks on EOG signal [9]

Figure 17: Effect of neck movement on EOG signal [9]

Some other sources of noise that can obstruct the EOG from being ideal are involuntary blinks and variations in the characteristics of skin-electrode interface. Both of these sources of noise can affect even a filtered signal.

Figure 18: EOG signals for various actions that include the eye motions and involuntary blinking

As shown (see figure 18), the involuntary blinking has about half the amplitude of the voluntary blinking, roughly 1.75V. The frequencies that the bandpass filters accepted ranged from 10 Hz to 150 Hz.

Finalized Design

After working on the individual part for a long time, all the parts have been put together at last. The linear actuator has been set in place and working perfectly. The rotation motor and the timing belt have been also installed, but the angle of rotation is a little limited due to the uneven radius of the chair’s base at different angles. At extreme angles the timing belt becomes too taut or too loose. Both relay circuits that control the motors had also been constructed last week. Now, it has been soldered to the necessary connections and made whole with the project. Although the code and the motors work perfectly fine individually, we have encountered some problems with running the entire assembly through EOG signals. We suspect an error either in the double relay circuits or the main power source. It may be that both the circuits are damaged and they are not reacting to the signals from the DAQ. It is more likely, however, that our power source cannot feed the entire circuit –only a motor at a time.

Figure 19: The entirety of the chair assembly merged together

Integration of the EOG Data

After recording the EOG signals in the MATLAB software, we will differentiate whether the individual is looking up and down or left and right. When the subject looks up or down, the motor in the linear actuator (built into the swivel chair) will move the seat correspondingly. If the individual looks up, the linear actuator will extend forcing the seat upwards with the disabled patient in the chair. If the individual looks down, the linear actuator will retract allowing the weight of the subject to push the chair slowly downwards. Next, we will have a motor controlled by the left and right eye movements. This motor will be connected to a timing belt which is wrapped around a swivel base of the chair allowing rotation up to a certain degree (to be established). Therefore, when the individual looks right, the motor will turn on in the corresponding direction, which will rotate a surrounding gear. This gear will then be attached to the timing belt resulting in the rotation of the timing belt which surrounds the base of the chair. This base will then rotate causing the rotation of the chair as a whole with the disabled person in it. The same system of operations occur when the subject looks to the left but in the opposite direction (see figure 15). Both the motor and the linear actuator will be powered by 12V of power which will be connected to a circuit allowing them to be switched on and off based on the collected EOG signals.

Figure 15: Flowchart depicting translation of EOG signals to mechanical aspects of chair design

EOG Signals

Raw EOG signals measure the electric signal from the user’s eyes, that measures all of electric signalling of the electrodes. Raw EOG signals capture both the voluntary and involuntary muscle contractions. The difference between raw EOG signals and ideal EOG signals is that ideal EOG signals filter out involuntary muscle contractions. Below are graphs comparing raw signals to filtered signals:

Figure 13: Graphs comparing raw EOG signals to filtered EOG signals (NEEDS REFERENCE)

The top two graphs (see figure 13) are raw signals that capture all signals with noise and involuntary contractions, while the bottom two graphs represent the positive data without the scattered, erratic noise.

By examining the ideal EOG signal shown below, we can determine the direction of an individual's eyes (see figure 14). When the patient looks either up or right, the wave gradually increases in the positive direction until it reaches about .1 mV (depending on the person), and then the wave drops back to the base line. When the individual looks either down or to the left, the wave gradually decreases in the negative direction until it reaches about -.1 mV and then rises again back to the base line. This ideal EOG signal is rarely obtained due to motion and noise artifacts (discussed in previous posts), however the general shape can be used to determine direction in any scenario (10).

Figure 14: Ideal EOG signal collected from left and right or up and down movement of eyes (10)

DAQ Testing and Plotting Signals

In lab this week, the amplifier and the electrodes were connected by soldering wires with alligator clips. Wires were also partially stripped in order to allow for a connection between the alligator clips, amplifier, and DAQ sensor. Using MATLAB, a code was produced to create a graphed stream of incoming DAQ signals. A filter was also encoded in the script to shield out noise and unwanted signals. Various plots were graphed when the script was activated, which included raw signals from the DAQ vs. time, filtered signals from the DAQ vs. time, and energy signals from the DAQ vs. time. The graphs of the raw signals vs. time and the filtered signals vs. time are visibly different; the raw signals are a lot more accurate but scattered, but the filtered signals have a much smoother, consistent graph.

Building the Mechanical Aspect of our Design

Two pieces of 3/4" plywood were cut into 24" diameter cylinders at the Drexel Hess lab, as shown in Figure 12.  Then, the wood pieces were glued and nailed together, and sanded on the sides. This 1 1/2" thick cylindrical base will be mounted on top of the swivel base.

The code first determines what devices are currently connected to the machine, and upon locating them, it creates a session to interface with a given vendor ID. Then, it sets the sampling rate to 48,000 and time information. Next, it sets the number of samples to be obtained or the duration (in seconds) for which the code must run.

Figure 12: Two circular pieces of swivel base

After receiving almost all parts of the mechanical design, a few pitfalls were encountered:

1.      The linear actuator, fully retracted, reaches from the top of the cylindrical base to the bottom of the chair. Simply, the linear actuator is too long This means that our team will have to mount the chair on top of a another box-like structure and put the linear actuator inside the box to make its base lower than the chairs base, as shown in Figure 13.

Figure 13: A possible redesign solution to the linear actuator height problem

2.      The gear motor, running on 12 volts, has very low torque. The swivel base purchased is very hard to turn, adding more concern to the low torque motor. A very high ratio gearbox will have to be purchased. The timing belt will be placed around this gear box, and then nailed to the cylindrical base. 

3.      The motor and gear box used to turn the swivel base must be mounted at the same height as the cylinders mounted on the swivel base. A simple extra box can be constructed to fix this problem.                         

4.      The timing belt may slip off of the cylindrical base; a guard must be constructed around the top and bottom of the cylinder holding the belt. This can be made out of the aluminum sheet we have already purchased.