Our brain communicates using electrical pulses -- from moving your hand to type on a keyboard to talking with your friend over the phone, all of your actions are mediated through a series of electrical pulses moving from your brain to the intended limb. Using electrodes, we can record these electrical signals and analyze them. For recording muscular activity, the resulting recording is known as an electromyogram (EMG). The EMG is of particular importance in clinical settings because it can help doctors diagnose certain diseases. If you have ever experienced muscle weakness, pain, or numbness, your doctor may have conducted a test to check if you have a problem with your nerves, spinal cord, or muscles. However, the EMG can also be of use to us in engineering systems for patients to use. Since EMG recordings are controllable, a potential application for the signal is communication

Biosignal-based communication is something that is currently studied extensively. EEG-based systems are used in patients in a locked-in state (can't move or talk) to predict intent in what are known as P300 systems. EMG systems are used less extensively, as the population that requires these systems have typically lost motor function. If they have not, typing is infinitely easier.

Nonetheless, the goal of this project is to design a system such that we can use EMG to communicate, as we did not have the equipment necessary for an EEG-based system.

The system must:

1. Communicate using a list of 20 phrases

2. Use only EMG signals

3. Be designed using LabVIEW

System performance will be evaluated using a performance index (PI), where PI = number of phrases successfully selected per minute - number of electrodes attached.

Data Acquisition

Data is acquired from the subject's forearm (three electrodes) at a 5k sampling rate and 2.5k samples to write using the DAQ block and then filtered using a bandpass filter from 0-500Hz. Though the frequency of EMG may extend far beyond this, most of the "usable" energy falls within this range. The signal is then converted into a numeric array using the dynamic data conversion block. Then, using a peak detector block, the maximum of these values is taken. This is then put into a makeshift comparator that compares this value to a threshold value, in this case 0.5 volts, and outputs a 0 or 1 depending on whether the recorded value is below or above this threshold.

At a high level, what we are doing is analyzing the EMG signal every time we record it and taking the maximum value. Then, we compare that maximum value to a threshold; if the value is below the threshold we output a 0, if it is above we output a 1. Basically what this does is detect whether or not a clench was actually completed.

Now we put this entire sequence into what is known as a shift register. Shift registers have the ability to run this sequence and store the value of the previous iteration. We use this to be able to store values into an array. We set this to run five times so we can have a five digit code at the end.

Array Conversion

Okay so now that we have our binary system, we want to put our values into an array. The array is initialized with the first value of the clench and we put more values in there until we reach our five sample limit. Then we have a converter that converts the number array into a decimal string. That string is combined using what is known as the concatenated string block, which combines all the values that we have into a five digit string. Then, we put our array into a MathScript node and assign phrases.

What this is doing is putting our values into an array one by one, converting it into a number format that our program can understand, and then putting it together. Then we take that string and assign it to each phrase.

LED Control

When we were testing the system we realized it was really hard to track what sample of the five-digit code you were on, so we decided to add some LEDs to help the user track what position the system was in. Basically we just added five LEDs, and made one turn on/off depending on how fast we were sampling the signal (in this case, it was 500 milliseconds).