# Characterization of capacitive electromyography biomedical sensor insulated with porous medical bandages

Sep 10, 2020

### Noise floor characterization

The noise floor characterization test measured the baseline result of the porous medical bandages insulated cEMG biomedical sensor. Since the EMG signal is a weak bioelectrical signal range from 0.05 to 5 mV, the noise floor of these sensors ought to be as low as possible to maximize the signal-to-noise ratio2. In this test, the subjects’ bicep brachii were required to stay idle and at ease during the measurement, to ensure no EMG signals were generated. Figure 4 shows the average peak-to-peak noise floor amplitude (({V}_{pk-pk})) measured from the two subjects. The micropore and the gauze recorded a similar noise amplitude of 2.44 mV and 2.47 mV. Crepe bandage recorded the highest noise floor of 7.386 mV.

The power spectrum analysis of the noise floor in Fig. 5 shows that the 50 Hz power line noise was the main contributor of the noise floor for all types of cEMG biomedical sensors. Micropore insulated cEMG biomedical sensor recorded − 65 dB, gauze insulated cEMG biomedical sensor recorded − 73 dB, and crepe bandage insulated cEMG biomedical sensor recorded − 50 dB. The crepe bandage also observed a harmonic frequency spike of − 86 dB at 150 Hz while other porous materials do not observe a harmonic frequency spike.

Since the experiment was done in an environment without electromagnetic wave shielding, the surrounding noises such as 50 Hz power line noise was expected to couple into the measurement results. Generally, a cEMG biomedical sensor has a high input impedance characteristic, thus it is highly susceptible to the environmental noises. The porous material that formed a lower skin–electrode capacitance will have a higher input impedance and eventually becomes the dominant factor of a cEMG biomedical sensor recording a higher noise floor. The 1-m cable connecting the cEMG biomedical sensor and the bandpass filter module is a single core ribbon cable insulated with PVC material. Since this cable is a non-shielded cable, thus it is also susceptible to the electromagnetic noise in the surrounding. This ribbon cable contains four parallel electrical lines which arrange in the sequence of + 15 V, − 15 V, analog ground, and the measured EMG signal. The EMG signal line is placed next to the analog ground and the experimental setup is powered by battery to reduce the interference of the 50 Hz power line noise to the EMG measurement results. Generally, a cable that is longer than λ/20 is consider electrically long and started to act as a reception antenna18. The Eq. (3) can be used to determine the wavelength of the highest frequency of concern.

$$lambda = frac{c}{f}$$

(3)

where λ is the wavelength in meter. ƒ is the target frequency in Hz. c is the speed of light which is (3 times {10}^{8} {rm {m/s}}).

For a 50 Hz noise signal, the value of λ/20 is 300 km. Therefore, a 50 Hz power line noise can hardly induce into the 1-m cable and corrupt the EMG signals. However, any high frequency noises which is more than 15 MHz is still possible to induce into the 1-m cable. The bandpass filter which has a passband frequency between 10 and 300 Hz will attenuate these high frequency noises in the EMG measurement results.

A 50 Hz digital notch filter was designed and implemented to attenuate the 50 Hz power line noise in the measurement to improve the overall noise floor of a porous medical bandage insulated cEMG biomedical sensor. The power spectrum analysis in Fig. 5 shows the significant reduction of 50 Hz noise in the baseline measurement after implemented the digital notch filter. Figure 4 shows that the 50 Hz digital notch filter has successfully reduced the noise floor of the micropore insulated cEMG biomedical sensor from 2.44 to 1.39 mV and gauze insulated cEMG biomedical sensor from 2.47 to 1.65 mV. The bandage insulated cEMG biomedical sensor has the greatest noise floor amplitude reduction from 7.39 to 2.28 mV. With the 50 Hz digital filter implemented, the difference of the noise floor amplitude among the porous materials has reduced from 4.94 to 0.89 mV, which makes them have a relatively similar baseline result.

### EMG burst signals measurement

The EMG burst signals measurement was to validate the capability of the porous medical bandage insulated cEMG biomedical sensor to measure a complete EMG burst signals from a muscle. Within a 10 s duration, each subject was required to perform dynamic contract of their bicep brachii for three times to generate the EMG burst signals. The amplitude of an EMG burst signals is expected to be around 0.05 mV to 5 mV2. All three types of the porous material were tested on each subject. Figures 6 and 7 present the raw and 50 Hz noise filtered EMG burst signals captured from subject A using different porous medical bandage as an insulator. Each EMG burst signal is around 1 s duration.

Since the micropore insulated cEMG biomedical sensor has the lowest noise floor, the three EMG burst signals were clearly observed in Fig. 6a. The gauze insulated cEMG biomedical sensor has higher noise floor compare to micropore insulated cEMG biomedical sensor, thus the EMG burst signals is not clearly seen in Fig. 6b. As for crepe bandage insulated cEMG biomedical sensor, the noise floor amplitude is much higher than the EMG signals, thus the frontend buffer is saturated and unable to observe the EMG burst signals in Fig. 6c. Without required post signal processing, micropore insulated cEMG biomedical sensor yield the best measurement result with the highest signal-to-noise ratio.

In order to improve the overall porous medical bandage insulated cEMG biomedical sensor’s performance, a 50 Hz digital notch filter was implemented in the post signal processing stage to reduce the power line noise in the measurement results. The 50 Hz digital notch filter has a 0.5 s settling transient during the start-up stage. The micropore insulated cEMG biomedical sensor with 50 Hz notch filter yielded a better signal-to-noise ratio measurement result as shown in Fig. 7a. All three EMG burst signals were clearly observed. The EMG burst signals measured by gauze insulated cEMG biomedical sensor also can be seen clearly after implementing the 50 Hz digital notch filter. The 50 Hz power line noise is greatly reduced in the crepe bandage insulated cEMG biomedical sensor measurement results after the implementing the 50 Hz digital notch filter. Three EMG burst signals can be observed in Fig. 7c.

### Performance evaluation

The performance evaluation validated the consistency, accuracy, and repeatability of the porous medical bandage insulated cEMG biomedical sensors in EMG signal measurement. The performance of these sensors was benchmarked with the conventional wet contact electrode (Ag–AgCl) which is a gold standard in clinical applications. In this experiment, all two subjects were required to contract their bicep brachii for 1 s to generate the EMG signal. The same EMG signal source were measured by both sensors. This test was repeated five times on every subject, to guarantee the repeatability and reliability of the measurement result.

Figure 8 shows the 1 s raw EMG signals captured by the micropore, gauze, and crepe bandage insulated cEMG biomedical sensor. EMG signals captured by the micropore and gauze insulated cEMG biomedical sensor have a similar signal pattern and amplitude compare to the EMG signals captured by the wet contact electrode. The crepe bandage insulated cEMG biomedical sensor was corrupted by the power line noise, therefore the signal captured is a 50 Hz periodical sine wave with uneven amplitude and it is completely different from the EMG signal measured by the wet contact electrode.

These raw EMG signals were post-processed by a 50 Hz digital notch filter to further evaluate the measurement results. Figure 9 shows the raw EMG signals in Fig. 8 filtered by the 50 Hz digital notch filter. Both micropore and gauze insulated cEMG biomedical sensor’s measurement results had a minimal improvement. Their EMG measurement results were still closely match with the EMG signals measured by the wet contact electrode. The EMG signals measured by the crepe bandage insulated cEMG biomedical sensor is no longer dominated by the power line noise (50 Hz periodical sine wave) after implemented the 50 Hz digital notch filter. Its pattern and amplitude were similar to the EMG measurement result captured by the wet contact electrode but not a close match.

The correlation coefficient (R) values of the EMG signals measured by the porous medical bandage insulated cEMG biomedical sensor and wet contact electrode were calculated. The average correlation coefficient values of the raw EMG signals of each subject were shown in Fig. 10. The correlation coefficient values were not expected to achieve 1.0 because both cEMG biomedical sensor and wet contact electrode were not possible to place at the exact same location. The micropore insulated cEMG biomedical sensor yielded a consistent measurement results between the subject A and subject B. High average correlation coefficient value of 0.83 and 0.84 were shown between the EMG signals measured by the wet contact electrode and the micropore insulated cEMG biomedical sensor. The EMG signals measured by the gauze insulated cEMG biomedical sensor have 0.84 and 0.71 correlation coefficient value with the wet contact electrode. The gauze insulated cEMG biomedical sensor yielded a relatively high correlation coefficient values, but the results were not consistent between the two subjects. The EMG signals measured by the crepe bandage insulated cEMG biomedical sensor and the wet contact electrode were uncorrelated which yielded the correlation coefficient values of 0.02 and 0.31 on subject A and subject B. Since the crepe bandage insulated cEMG biomedical sensor was corrupted by the power line noise as shown in Figs. 6 and 8, an uncorrelated result was expected.

The correlation coefficient value of the EMG signals post-processed by the 50 Hz digital notch filter were calculated as well to evaluate the performance improvement. The average correlation coefficient results of the EMG signals post-processed by the 50 Hz digital notch filter were shown in Fig. 11. The EMG signals measured by the micropore insulated cEMG biomedical sensor and the wet contact electrode yielded the highest correlation coefficient value of 0.84 and 0.85 on subject A and subject B. Since the micropore insulated cEMG biomedical sensor originally has a relatively low noise floor, the 50 Hz digital notch filter does not bring a big improvement to the measurement results. The EMG signals measured by the gauze insulated cEMG biomedical sensor and the wet contact yielded the correlation coefficient value of 0.85 and 0.77 on subject A and subject B. The 50 Hz digital notch filter improved the correlation result of subject A by 0.01 and subject B by 0.06. The performance of a crepe bandage insulated cEMG biomedical sensor had the greatest improvement after the implementation of a 50 digital notch filter. The correlation coefficient of the EMG signals measured from subject A improved from 0.02 to 0.41 while subject B improved from 0.31 to 0.67. The 50 Hz digital notch filter was successfully attenuated the power line noise in the signal measurement.

Overall, the porous medical bandages were suitable to be used as an insulator of a cEMG biomedical sensor while different bandages will yield a different performance. A high input impedance cEMG biomedical sensor is highly susceptible to the environmental noise. A porous medical bandage insulated cEMG biomedical sensor which yielded a lower noise floor would eventually achieve a better EMG measurement result due to a better signal-to-noise ratio. A 50 Hz digital notch filter is proven to be able to effectively attenuate the dominant power line noise coupled into the cEMG biomedical sensor and improved the sensitivity of the sensor.