Open Access
Add to BibSonomyAbstract
A wearable electroencephalogram (EEG) is a small mobile device used for long-term brain monitoring systems. Applications of these systems include fatigue monitoring, mental/emotional monitoring, and brain–computer interfaces. However, the usage of wireless wearable EEG systems is limited due to the risks posed by the wireless RF communication radiation in a long-term exposure to the human brain. A novel microwave radiation-free system was developed by integrating visible light communication technology into a wearable EEG device. In this work, we investigated the system’s performance in transmitting EEG data at different illuminance level and proposed an algorithm that functions at low illuminance levels for increased transmission distance. Using a 30 Hz smartphone camera, the proposed system was able to transmit 2.4 kbps of error-free EEG data up to 4 meter, which is equal to ~300 lux using an aspheric focus lens.
© 2017 Optical Society of America
1. Introduction
Electroencephalogram (EEG) monitoring is an electrophysiological monitoring method to record the electrical activity of the brain. Conventional EEG acquisition tools are equipped with many electrodes and involve complex measurement preparation methods. However, wearable EEG headsets are a better alternative as they require fewer electrodes and are simpler to prepare; they also have enhanced wearability and mobility for long-term healthcare monitoring. Most EEG devices use wireless RF communication technologies such as Bluetooth and ZigBee for wireless data transmission. However, Bluetooth and ZigBee devices emit microwave radiation that may cause biological changes in the human body [1]. Long-term exposure to microwave signals from Bluetooth devices may cause health problems such as brain cancer, brain tumor, stress, and leukemia. The electromagnetic interference (EMI) from RF communication also creates noise that damages EEG signal data. One solution to these problems is the use of visible light communication (VLC) in wearable EEG monitoring systems, which is introduced in this study.
Recently, VLC systems have attracted much attention as next-generation communication systems. The advantages of using VLC compared to other common RF communication technologies are that it is safe for the human body, capable of high-speed data communications, provides secure communication, and is EMI-free, which is important for healthcare data communication. Various studies have been conducted on the development of VLC for real-life applications. Researchers proposed the combination of LEDs and solar cells for a VLC and energy harvesting system [2,3], while others implemented VLC in environmental monitoring [4], vehicle-to-everything [5], indoor positioning [6], and health care data transmission. The proposed VLC is defined by the communication between light from the LED and the complementary metal-oxide-semiconductor (CMOS) image sensor of a smartphone camera. In this particular research area, modulation methods have been developed by some researchers, including [7], Christos et al. utilizing the effect of a rolling shutter from a built-in CMOS smartphone camera sensor array. However, their method was restricted by the short communication distance (<30 cm). Other methods called under-sampled frequency shift on-off keying (USFOOK) [8], and under-sampled phase on-off keying (USPOOK) resolved the distance problem [9], attaining a communication distance of 12 m, but these methods had problems with data transmission speed.
In this study, we propose a novel VLC-based wearable EEG to serve as a convenient, long distance mobile healthcare monitoring system with a radiation-free communications module. We focus on developing a system with a demodulation algorithm that can decode OOK signals from a lower illuminance-received image frame (long distance). The experimental results show that the proposed system was able to communicate without bit errors at ~300 lux, which is two times better compared to other algorithm [10–13]. The proposed demodulation algorithm including data packet design, data synchronization, and image frame processing are discussed in detail in this study. Bit error rate (BER) evaluation and the relation between measurement distance and illumination are also discussed theoretically and experimentally to validate the results of our system. The discussion is restricted to these topics, as digital processing techniques such as feature extraction and machine learning classification are significant and independent topics that will be covered in future publications.
2. System structure
The proposed wearable EEG consists of electrodes and a microcontroller (MCU) for the analog data filter and analog to digital converter (ADC). A single-channel EEG was designed to collect brain activity data from the user’s brain. Simplicity and efficiency were the main reasons for designing a single-channel EEG monitoring system. Simplicity was achieved by using only three electrodes as a minimum requirement to receive EEG data. The results shown in [14,15], prove that a single-channel wearable EEG is sufficient and useful in certain monitoring applications. For example, placing electrodes over the occipital regions may enable a system to record an activity related to eyes that are open and closed. On the other hand, placing electrodes over the central regions may enable ca system to record an activity related to movement. As seen in Fig. 1, right after the electrodes sensed the data, the VLC with LED transmitted the data to the smartphone through the light.
Fig. 1 Proposed system block diagram showing EEG signal acquisition, the LED as transmitter, and the smartphone as receiver.
The real-time measurement setup is shown in Fig. 2(a). As seen in the figure, we use optical table and an adjustable tabletop smartphone tripod to prevent any vibration that potentially corrupt the result of the experiment. Furthermore, the proposed wearable EEG consists of three electrodes, designated as 2 EEG input electrodes, and ground electrode, as seen in Fig. 2(b). It is powered by a 3.7 V 2000-mAh lithium-ion battery. The EEG signals from electrodes were converted to digital data using a 16-bit ADC. In order to obtain the most useful EEG bands that were Ɵ (theta) waves of 4–7 Hz, α (alpha) waves of 8–13 Hz, and β (beta) waves of 13–30 Hz, a 4 Hz high-pass filter and 32 Hz low-pass filter were chosen. The high-pass and low-pass filters determined the minimum sampling rate of the system. According to the Nyquist sampling law, sampling rate is determined by the frequency region of interest (ROI) times two. Thus, if the frequency ROI is 0–50 Hz, the minimum sampling rate should be 100 Hz. The proposed system was aimed at monitoring multiple applications as seen in Figs. 2(b); different placements of the electrode may enable the system to receive multiple brain activity signals. For example, the electrodes were placed in the occipital regions O1 and O2, which were highly correlated with the driver’s vigilance level.
Fig. 2 (a) Wearable EEG measurement setup (side look) and (b) front look, with different electode placement.
3. Visible light signal processing algorithm
A high data rate is required to transmit EEG data wirelessly over the proposed system. A clinical EEG is commonly recorded at sampling rates of 250 Hz or 500 Hz. However, for wearable EEG, sampling rates vary depending on the purpose of the system. Thus, a 150 Hz sampling rate was chosen for the proposed system [15]. Unfortunately, embedded cameras in most smartphones are limited to 30 Hz, making it impossible to communicate with a wearable EEG. The rolling shutter effect was utilized in the proposed system to enable high-speed EEG wireless communication. Most of the CMOS image sensors embedded in the smartphone were operated using the rolling shutter. Unlike the global shutter that captures the entire image all at once, the rolling shutter activated each row of pixels in the sensors sequentially, which meant that if the data transmission rate was much higher than the camera sampling rate, bright and dark lines could be recorded in one image. Bright lines appeared if the pixel rows were exposed by the LED light, if not dark lines appeared.
As shown in Fig. 3 (a), the proposed algorithm is divided into two parts, the initial image processing and demodulation algorithm. In contrast to algorithms proposed in [10,11], our proposed algorithm provides a simple image thresholding and intensity adjusting method, yet has better performance at lower illuminance. It starts by reading the image frame through the back-camera of the smartphone. When captured by the image frame, the image is cropped to a 300p width, 150p to the left and 150p to the right from the middle of the image. Our experiments showed that a 300p wide image was compulsory to obtain accurate raw data. Accordingly, we adjusted the intensity of the received image and found that the distinguishing data appeared between 50% and 70% of the maximum received grayscale value. Thus, the cropped adjusted-image was converted into grayscale and the column selected by averaging the value of the 300 horizontal pixel row as a single-column grayscale value matrix. The selected column represented the data horizontally. In order to find the data header, the data threshold was first applied to the selected column. The data header consisted of a 4-bit periodic signal encoded using Manchester coding. We also applied a second data threshold to some special frames for which the system failed to detect the header. This second data threshold was only used to find the data header; for data payload thresholding, we still used our first data threshold. The output of this data threshold was bit “0” and “1”. Thus, we visualize the output data after thresholding as an image and compare it with the received image before the proposed algorithm as shown in Fig. 3 (b). Based on our experiment, the best performance values for the first and second thresholds were achieved by dividing the maximum received grayscale values by 2.3 and 4, respectively. Furthermore, as 1 bit data consisted of multiple pixel columns, we proposed a window slide-based decoding method to classify the received data payload. Our proposed window slide-based decoding method was also useful in mitigating error detection due to the blooming effect. As mentioned in [10,13], the blooming effect created different ratios between bright and dark lines, where dark lines had a narrower periodic compare to bright lines. As we achieved 6.375 pixels/bit, our slide window length was also a 6.375 pixel column. Then, bit output decision-making was collected from the maximum value between columns 3 and 4 of the corresponding window.
Fig. 3 (a) Rolling shutter-OOK demodulation alghorithm (b) Example of received image before and after proposed processing algorithm.
4. Result and discussion
Several experiments were conducted to examine system performance. The BER was checked to compare Bluetooth technology and VLC transmitting the same payload. Different distances were also checked to determine the maximum system performance for bit-error-free communication. A single 3W RGB LED Common Anode (Shenzhen Fedy Technology Co., Ltd., China, FD-3RGB-Y2) transmitter was used to transmit OOK-modulated green color, while a 75 mm diameter × 50 mm focal length aspheric lens (Edmund Optics Korea, Ltd., Korea) was used to focus the light. A Samsung Galaxy S6 was used as the receiver. The minimum requirement for the smartphone camera used in this experiment was Camera 2 API (Android 5.0 APIs) to control the manual setting of the camera, as given in Table 1. Table 1 lists the specific parameters used in the experiment. By using those parameters, we maximized the potential of 1920 × 1080 resolutions with a 30 Hz camera. Initially, similar to [10], we needed to implement the oversampling method to prevent packet losses due to the frame to frame time gap. Thus, we sent each payload data redundantly, as shown in Fig. 4; Five packets of EEG data were transmitted all at once; two headers can be seen in each frame and the payload contained 80 bits of data (1 EEG data = 16 bits).
Table 1. Smartphone Camera Parameter
Fig. 4 Data Packet structure with 2 possible received packet (data1 and data2 represented redundant payload).
Figure 5 shows the performance of the proposed algorithm at different illuminance levels. As seen in the top parts of the figure, the received grayscale value was reduced as the illuminance level decreased and weakened the detection performance of the system. Our proposed algorithm was able to overcome these problems, as seen in the bottom part of the figure. Our experiment shows that most of the errors were caused by the failure of the system to detect the data header, since it is a very thin but varied grayscale value bright lines. By adjusting the intensity of the received image as mentioned in the previous section, the system remained free of bit errors until 327.84 lux or equal to a 450 cm distance with an aspheric lens. In addition, we also test the relation between vibration and the received data by holding the smartphone with hand while recording the signal. The results shows that the small vibration comes from human body did not affect the results of our received data.
Fig. 5 Normalized grayscale value of the received image at different light illuminance: (a) 259.54 lux, (b) 327.84, and (c) 464.44 lux.
The overall BER performance of the proposed system can be seen in Fig. 6(a). Furthermore, to check the validity of the illuminance after the aspheric lens, we compared the perceived illuminance with the inverse square law. The results in Fig. 6(b) show the relationship between illuminance and distance. The results prove that the perceived light illuminance was comparable with related theoretical data. Regardless of the lens used, a comparison of our proposed system with other demodulation algorithms showed that our proposed system exhibited good performance with two times lower illuminance. This means that, two times lower illuminance is equal to lengthen the distance by 1 meter, based on the inverse square law and our results. We believe that this 1 meter additional distance can be useful for certain applications—in this case, healthcare applications. Finally, a comparison of the wearable EEG data transmission of the proposed system with common Bluetooth communication is shown in Figs. 6(c) and 6(d). As seen in the figure, we were able to achieve identical data without bit errors in the 450 cm measurement distance.
Fig. 6 System performance in (a) different illuminance (b) comparison with theoretical data in different distance, and (c) comparison Bluetooth communication.
5. Conclusion
This study proposed mobile LED VLC communication as a microwave radiation-free data transmission protocol for a single-channel wearable EEG device. An LED was used as the transmitter and the CMOS image sensor module of a smartphone camera was used as the receiver in the wearable EEG monitoring system. Rolling shutter-effect detection served as an algorithm to identify received light, and multiple experiments with different distances were conducted to determine the system’s robustness. We proposed a demodulation algorithm composed of two-step processing with image thresholding and intensity adjusting. The proposed algorithm had good performance at lower illuminance level. Thus, by utilizing a focus lens, the proposed system was able to communicate without packet loss up to a distance of 450 cm or equal to 327.83 lux. Furthermore, in order to prove that the proposed alternative communication method can replace current Bluetooth technology, the performance of both systems was tested and compared by sending the same payload for the wearable EEG monitoring system. The proposed system was able to send 2.4 kbps of error-free data up to a transmission distance of 450 cm.
Funding
This work (2016R1A2B4015) was supported by Mid-Career Researcher Program through an NRF grant funded by (MEST), Korea.
References and links
1. H. Lai, “Neurological Effects of Non-Ionizing Electromagnetic Fields,” http://www.bioinitiative.org/report/wp-content/uploads/pdfs/sec09_2012_Evidence_%20Effects_%20%20Neurology_behavior.pdf.
2. Y. Liu, H. Y. Chen, K. Liang, C. W. Hsu, C. W. Chow, and C. H. Yeh, “Visible Light Communication Using Receivers of Camera Image Sensor and Solar Cell,” IEEE Photonics J. 8(1), 1–7 (2016).
3. S. H. Lee, “A passive transponder for visible light identification using a solar cell,” IEEE Sens. J. 15(10), 5398–5403 (2015). [CrossRef]
4. Z. Ong and W. Y. Chung, “Long range VLC temperature monitoring system using CMOS of mobile device camera,” IEEE Sens. J. 16(6), 1508–1509 (2016). [CrossRef]
5. Y. Goto, I. Takai, T. Yamazato, H. Okada, T. Fujii, S. Kawahito, S. Arai, T. Yendo, and K. Kamakura, “A New Automotive VLC System Using Optical Communication Image Sensor,” IEEE Photonics J. 8(3), 1–17 (2016). [CrossRef]
6. Y. Nakazawa, H. Makino, K. Nishimori, D. Wakatsuki, and H. Komagata, “Indoor positioning using a high-speed, fish-eye lens-equipped camera in visible light communication,” in Proceedings IEEE Indoor Positioning and Indoor Navigation (IPIN, 2013), pp. 1–8. [CrossRef]
7. C. Danakis, M. Afgani, G. Povey, I. Underwood, and H. Haas, “Using CMOS Camera Sensor for Visible Light Communication,” in Proceedings IEEE Globecom Workshops (GC Wkshps, 2012), pp. 1244–1248. [CrossRef]
8. R. D. Roberts, “Undersampled Frequency Shift ON-OFF Keying (UFSOOK) for Camera Communications (CamCom),” in Proceedings 22nd Wireless and Opt. Comm. Conference (WOCC, 2013), pp. 645–648. [CrossRef]
9. P. Luo, Z. Ghassemlooy, H. L. Minh, X. Tang, and H. M. Tsai, “Undersampled phase shift ON-OF keying for camera communication,” in Proceedings IEEE Wireless Communications and Signal Processing (WCSP, 2014), pp. 1–6. [CrossRef]
10. C. W. Chow, C. Y. Chen, and S. H. Chen, “Visible light communication using mobile-phone camera with data rate higher than frame rate,” Opt. Express 23(20), 26080–26085 (2015). [CrossRef] [PubMed]
11. Y. Liu, C. W. Chow, K. Liang, H. Y. Chen, C. W. Hsu, C. Y. Chen, and S. H. Chen, “Comparison of thresholding schemes for visible light communication using mobile-phone image sensor,” Opt. Express 24(3), 1973–1978 (2016). [CrossRef] [PubMed]
12. K. Liang, C. W. Chow, Y. Liu, and C. H. Yeh, “Thresholding schemes for visible light communications with CMOS camera using entropy-based algorithms,” Opt. Express 24(22), 25641–25646 (2016). [CrossRef] [PubMed]
13. K. Liang, C. W. Chow, and Y. Liu, “RGB visible light communication using mobile-phone camera and multi-input multi-output,” Opt. Express 24(9), 9383–9388 (2016). [CrossRef] [PubMed]
14. L. W. Ko, W. K. Lai, W. G. Liang, C. H. Chuang, S. W. Lu, Y. C. Lu, T. Y. Hsiung, H. H. Wu, and C. T. Lin, “Single channel wireless EEG device for real-time fatigue level detection,” in Proceedings IEEE International Joint Conference on Neural Networks (IJCNN, 2015), pp. 1–5.
15. G. Li, B. L. Lee, and W. Y. Chung, “Smartwatch-based wearable EEG system for driver drowsiness detection,” IEEE Sens. J. 15(12), 7169–7180 (2015). [CrossRef]
