[0:01]Hi, my name is Joshua Silver and I'm here to talk about EMG signal processing and analysis.
[0:14]So what is EMG? Well, EMG stands for electromyograph, and it's used in the medical field for measuring electrical impulses from muscular activity. It commonly has a frequency of between 1 and 10 millivolts and a frequency ranging from 0 to 500 Hz, although it's more commonly found between 50 and 150 Hz. This is different than other biopotentials, which are various electrical impulses that our bodies gives off, such as ECG, which is which measures electrical impulses given off from the heart or EEG, which is impulses from the brain. Here I have outlined the general process at which a person could take a surface electrode to measure these electrical impulses from a patient's muscle, and then process that raw EMG data into data that can be analyzed by a physician to detect neurodegenerative diseases or be used in other applications.
[1:18]Like all biopotentials, EMG signals have a very low amplitude. So we generally use a instrumentation amplifier, which gives off low noise of a gain of about 100 to give us a larger signal that is easier to analyze and understand. This is generally implemented in hardware.
[1:42]Since this electrode that we are using can measures all sorts of electrical potentials, it's important to include an anti-aliasing filter. This is generally a low pass filter of that has an upper end of around 1 kHz. This is going to filter out our unwanted signals. We choose a 1 kHz to satisfy the Nyquist sampling theorem as that's about two times the maximum frequency that we will see from an EMG signal. This is again implemented in hardware.
[2:17]Once we have filtered our data, we can then put it into an analog to digital converter using a microcontroller. Generally in the professional field, they use 16-bit sampling, but it can also be used as low as 8 bits. Once we have this data sampled, the easiest way that I have found would be to transfer this data set to be analyzed in a program such as Python or MATLAB, where we can understand our data better than if we were to try to perform this signal processing with our embedded system.
[3:00]Once we have obtained our set of data from the ADC, we can then take the fast Fourier transform to bring us from the time domain to the frequency domain. Within the frequency domain, we can find the high and low cutoff frequencies of our signal. Using these frequencies, we can create a bandpass filter. This is necessary as we have high frequency noise given off from imperfection of the sensor and low frequency noise that is generated from the electrode shifting on the patient's muscle. The formula for the frequency response of an nth-order Butterworth filter is shown below, and using our cutoff frequency, we can define the pass band width W and the center frequency F0. We can then multiply this function with our data in the frequency domain before taking the IFT to bring us back to the time domain.
[3:59]Next, we have to do full wave rectification on our signal as processed EMG data is only represented in positive amplitudes. In hardware, this could be done using a bridge rectifier, but since we are working in software, it is easier just to take the absolute value of the negative components of our signal and add it to the positive components.
[4:28]After rectification, the data is usable, but it is better to both smooth and average our data to make it easier to understand. This data can then be analyzed to be used to determine various neurodegenerative diseases, such as Parkinson's or essential tremor.
[5:01]So what does this process do to our signal step-by-step? Well, in green, we can view the raw data, which would be coming out of the ADC. In red, we show our filtered signal, which comes out of our bandpass filter that we designed in software. Following that, we have our full wave rectification in blue. Following that, we have our smoothing, which is eliminating a lot of unnecessary noise and averaging, which is giving us a nice clean final signal.
[5:40]Thank you for listening to my presentation. Here I have the references that I used, and I hope you have a good day and happy holidays.
