Development of a wireless wearable electrooculogram recorder for IoT based applications Suvodip Chakraborty1 , Anirban Dasgupta2 , Punyashlok Dash3 and Aurobinda Routray2 Abstract— Internet of things (IoT) has captured a promising market in industrial electronics. Considering the scenario, we introduce a wearable wireless electrooculogram (EOG) recorder for IoT-based industrial applications. This device has advantages over existing EOG recorders concerning its ease of wearing, portability and usability. The recording software is an Android app which makes it useful for a common man as well as many researchers working on eye-tracking based IoT applications. The system consists of Ag plated Cu electrodes for capturing the bio-potential near the canthus and the forehead. The analog EOG signal is pre-amplified using a signal conditioning circuit, comprising of an instrumentation amplifier, a bandpass filter and a differential amplifier. An embedded Wi-Fi module is used for transmission of data. The system has been compared with standard EOG recorders, and the results show it has comparable SNR and sampling rates with the existing recorders. Index Terms— EOG; Android; Wi-Fi
placed around the eyes [9]. Two of them are for the horizontal eye movement of each eye while the remaining two measure the vertical [10]. The reference electrode is usually located in the center of the forehead. Horizontal movement of eyeballs are captured by horizontal electrodes while the vertical electrodes capture blink and vertical motion of eyeball.The amplitude of the captured EOG signal depends on many factors as reported in [3]. They are: • perturbations caused by other sources(e.g. facial muscles) such as the electroencephalogram (EEG), electromyogram (EMG) • the acquisition system • positioning of the electrodes • head movements, blinking
I. I NTRODUCTION Internet of Things (IoT) has dominated the area of Industrial Electronics in the past three years. Recently, biomedical signals are used in many IoT-based applications, which range from medical diagnosis, health care to the assessment of cognitive and affective states. One such biosignal is the electrooculogram (EOG). EOG is a measure of the resting potential between the cornea and the retina of the human eye [1]. This potential exists due to the hyperpolarisations and dehyperpolarisations of the neurons when the eye moves [2]. The amplitude of the EOG signal varies from 50 to 3500 μ V with a bandwidth of100 Hz approximately [3]. EOG plays a key role in human behavior analysis, patient monitoring and medical diagnosis as reported in [4]. Barea et al. [3] describe a wheelchair for the differently-abled people using EOG signals. Venkataramanan et al. [5] have developed a hospital alarm system using EOG. Chieh et al. [6] have designed a drowsiness detection system for vehicle drivers using EOG. Dhillon et al. [7] have developed a Virtual Keyboard which accepts eye-gaze controlled navigation using EOG. Vidal et al. [8] have devised a method for activity recognition based on classifying eye movements from EOG. Thus, EOG signals finds a lot of significance in Biomedical and Industrial Electronics. The EOG signals are captured usually using five electrodes *This work was not supported by any organization 1 S. Chakraborty is with the Advanced Technology Development Center, Indian Institute of Technology Kharagpur, India 721302
[email protected]
2 A. Dasgupta and A. Routray are with the Department of Electrical Engineering, Indian Institute of Technology Kharagpur, India 721302 P. Dash is with the Department of Computer Science and Engineering, Indian Institute of Technology Kharagpur, India 721302
l-)))
Fig. 1.
Different Views of the Headgear based EOG recorder
The commercially available EOG recorder from Qubit, named S225 [11] use C410 LabPro Interface and C901 Logger Pro software for data acquisition which makes its use very precise. They provide a reset button which allows the instrument to be zeroed when the subject looks straight ahead. S225 has the capacity to accommodate up-to six electrodes. The units available from Biopac are MP36R, MP160, Mobita versions. MP36R has four channels of data acquisition and is capable of transferring the data to a windows based computing platform for extraction and computation of data acquired. MP160 is an upgraded version of the MP160. It is Ethernet-ready data acquisition and analysis that is capable
of recording data at different sampling rates. It boasts of a maximum sampling rate of 4KSPS. Mobita has an onboard storage and wireless transmission capability these devices however are not solely EOG acquisition device these devices can be used to record ECG, EEG, EGG and EMG. The S225, MP36R, and MP160 are wired type systems and hence constrain the movement of the user, during the capture. Mobita is a wireless recorder, however, only support a limited set of devices. This necessitates the development of a wireless EOG recorder, which is portable, easy to wear and also compatible with Android, to capture a wider audience. Considering the scenario, we have developed the EOG recording system, where the acquisition unit is in the form of a headgear, while the software is in the form of an Android app. The headgear developed is easy to wear, as compared to existing electrode systems, which require placing of electrodes individually. Fig. 1 shows a couple of views of our designed prototype headgear. The hardware and software development are treated in the following sections. II. T HE HARDWARE DESIGN As discussed earlier, our proposed design embeds the sensors along with the associated circuitry in a headgear, to make a compact design. The system hardware comprises of the electrodes, a signal conditioning circuit, power management circuit, batteries, the headgear and a charging socket. The positioning of the acquisition unit and its subcomponents is shown in Fig. 3. The components of the prototype system are described as follows. A. The EOG electrodes The electrodes comprise of dry Ag plated Cu plate sensors. We use a three-electrode system to capture the horizontal movements of the eye. This is carried out for two reasons: • horizontal eye movements can solely determine the cognitive and affective states of a person [12]. Therefore, only horizontal eye movements were considered for this work. • less number of electrodes increase compactness and decrease the hardware complexity and cost • vertical electrodes capture blinks as artifacts which makes them difficult to isolate The headgear has two electrodes placed near the canthus of each eye capturing the movement of the corresponding eye. A third electrode serves as the reference, and is placed at the centre of the forehead. The sensors capture the standing potential between the cornea (side electrode) and the retina (reference electrode). B. Signal Conditioning Circuit The acquired EOG signal is filtered and amplified by a signal conditioning circuit which comprises of an instrumentation amplifier , a bandpass filter and a differential amplifier. The instrumentation amplifier used is AD620 with a CMRR of 110dB. The bandpass filter is a RC Besselworth filter of third order, having a cut-off frequency of 30 Hz and voltage gain of 1000. The Fig.5 the detailed description of each and
every component with respective schematics are discussed in details C. Wireless Module The wireless transmission is governed by an embedded Wi-Fi module, ESP8266. This module transmits the EOG data from the analog circuitry to an Android device. This module maintains a synchronous connection with an Android device and transmits data when the device sends a ping. The above mentioned module has a low power 32-bit microcontroller and a built-in 10-bit ADC operating at 3.3 V. The sampling rate is set at 200 samples per second, which has been empirically found to be optimum as a trade-off between power consumption and speed of operation. An additional buffer of 400 samples is provided in the module to subsist the problem of mis-communication. D. Power management circuitry The power management circuitry consists of monolithic power switching regulators optimized for applications where power drain minimization is of essence. The device mainly operates as a variable frequency, voltage mode boost regulators and is designed to operate in continuous conduction mode. The low noise variable frequency voltage-mode DCDC converters consist of Soft-Start circuit, feedback resistor, reference voltage, oscillator, PFM comparator, PFM control circuit, current limit circuit and power switch. Due to the onchip feedback resistor network, the system designer can get the regulated output voltage from 1.8 V to 5 V. The leading features of our developed signal conditioning circuitry include: • •
Quick switching between sleep/wake patterns of 3 μs Low-power operation adaptive radio bias of 30 mW
The circuitry is designed with minimal necessary signal conditioning functions. The prototype board is light-weight, battery-powered, low power consuming device and is wireless enabled for sending acquired data via the ESP 8266 Wi-Fi module. This prototype board provides high-precision EOG signal with a less than 2.0 μV peak-to-peak noise. The board supports a sampling rate of up to 1000 Hz. The transmission module is kept alive for transfer for 0.00061 seconds and the rest of the time in sleep mode. In deep sleep mode, the ESP8266 maintains its real-time clock but shuts everything else off to hit about 60 mA. It can pull upwards of 200 mA while its transmitting, and an average of about 75 mA in normal operation. In deep-sleep mode the current consumption comes down to about 77μA. The positioning of the acquisition unit and its sub-components are shown in Fig.4. E. Batteries The wearable device runs on two Li-ion batteries that are rated at 150 mAh. The headset has an on-off power switch that is connected to an LED to ensure that the device is switched on.
III. T HE S OFTWARE D EVELOPMENT A. Processing unit The processing unit used in our prototype development is an ARM cortex-A based SoC residing in an Android Smartphone. The smartphone has an ARM-based dual quadcore Exynos chipset with the higher power cores clocking at 1.9 GHz and the lower at 1.3 GHz, with 3 GB RAM. The Android operating system runs on API level 21 version 5.0 (i.e. Lollipop), with back-compatibility up to 4.0 (i.e. Ice-cream Sandwich). B. Android Application Fig. 2. A sample case of EOG data as received on the smartphone alongwith filtering where a left eye movement is characterized by a positive signature and vice-versa. The red graph is the final filtered signal.
F. Headgear The headgear is made up of perspex acrylic sheets, and has cut-outs to support placement over ears. G. Charging socket The charging socket is a mini-USB socket, similar to that of a smart-phone charging unit. .
The front end of the application is shown in Fig. 6. Here the user has to enter the IP address and port of the TCP host for communication to be established. The IP and port address are pre-programmed in the ESP module which acts as the TCP host. Connect : The user then presses the connect button to connect to the Wi-Fi module. Upon successful connection a pop-up message appears confirming the same. In cases of failure, the pop-up notifies a cause for possible failure. The data is in byte-stream format and is written live onto a file named ’LiveData.txt’ present inside the internal storage of the phone upon successful connection. Disconnect : The user can disconnect the TCP connection by pressing the disconnect button. The disconnect button closes the TCP connection and stops writing to the file. Report : After closing the connection, the user needs to press the report button to generate a report of his/her performance. The app now uses the raw EOG data file i.e. ’LiveData.txt’ and first performs FIR filtering and then Kalman filter to remove noise. The report button can also plot the signal on the android device as shown in Fig.2. IV. R ESULTS
Fig. 3. system
Placement of electrodes and Wi-Fi module as embedded in the
We compare the Signal-to-noise ratio (SNR) to assess the performance of the developed system along-with the existing systems. The SNR is the ratio of powers present in the signal to that of the noise, and is usually expressed in decibels. Since, we do not exactly know the noise content, we carry out a blind estimation of SNR. Edfors et al. [13] have reported that Spectral Decomposition provides a reasonable blind estimate of SNR, as noise occupy the lower eigenvalues of the co-variance matrix. The SNR estimate is obtained as SNR =
2 σsignal . 2 σnoise
(1)
2 2 Here, σsignal and σnoise gives the largest singular value of the signal and noise respectively.
V. C ONCLUSION
Fig. 4. The components of the acquisition unit (a) batteries (b) electrode (c) the power management and signal conditioning PCB’s
We propose a wireless wearable EOG recorder in the form of headgear. The main advantages of our system over existing devices are its ease of use, compactness and portability. The novelty of the scheme lies in the design of the headgear to accommodate not only the electrodes but also the signal conditioning and power management units
Fig. 5.
Acquisition circuit on wearable system TABLE I
COMPARISON WITH EXISTING DEVICES
Manufacturer Qubit MP160 MP36R Mobita Developed
Sampling Rate (Hz) 250 200 200 200 250
Conductive medium Conductive pads Ag/Agcl Ag/Agcl Ag/Agcl Dry
Transmission of data USB USB/Wireless USB/Wireless USB/Wireless wireless
SNR (db) 5.2 8.9 8.2 8.6 7.8
Platforms supported Windows/Mac Windows/Mac Windows/Mac Windows/Mac Android
R EFERENCES
Fig. 6.
Home screen of the main app
on a PCB, batteries and allied connections. The system has comparable SNR’s and data sampling rates as compared to existing wired systems. The major hindrances in the development of industrial IoT-based applications using our proposed wireless EOG recorder are the requirement of IoT experts, optometrists and industry partners coming together. VI. F UTURE W ORK The major target of this system is the IoT, where EOG signals may be used for various applications ranging from cognitive studies to the development of industrial-grade health care devices. We plan to implement a cloud based system in the future where EOG data will be uploaded from multiple users and its data can be analyzed online and be used for monitoring various patients and drivers.
[1] A. Chaudhuri, A. Dasgupta, and A. Routray, “Video & EOG based investigation of pure saccades in human subjects,” in Intelligent Human Computer Interaction (IHCI), 2012 4th International Conference on. IEEE, 2012, pp. 1–6. [2] A. Chaudhuri, A. Dasgupta, S. Chakrborty, and A. Routry, “A lowcost wearble portable eog recording system,” in Proceedings of IEEE International Conference on Systems in Medicine and Biology, 2016. [3] R. Barea, L. Boquete, S. Ortega, E. L´opez, and J. Rodr´ıguez-Ascariz, “Eog-based eye movements codification for human computer interaction,” Expert Systems with Applications, vol. 39, no. 3, pp. 2677–2683, 2012. [4] D. Kumar and E. Poole, “Classification of eog for human computer interface,” in Engineering in Medicine and Biology, 2002. 24th Annual Conference and the Annual Fall Meeting of the Biomedical Engineering Society EMBS/BMES Conference, 2002. Proceedings of the Second Joint, vol. 1. IEEE, 2002, pp. 64–67. [5] S. Venkataramanan, P. Prabhat, S. R. Choudhury, H. B. Nemade, and J. Sahambi, “Biomedical instrumentation based on electrooculogram (eog) signal processing and application to a hospital alarm system,” in Proceedings of 2005 International Conference on Intelligent Sensing and Information Processing, 2005. IEEE, 2005, pp. 535–540. [6] T. C. Chieh, M. M. Mustafa, A. Hussain, S. F. Hendi, and B. Y. Majlis, “Development of vehicle driver drowsiness detection system using electrooculogram (eog),” in Computers, Communications, & Signal Processing with Special Track on Biomedical Engineering, 2005. CCSP 2005. 1st International Conference on. IEEE, 2005, pp. 165–168. [7] H. S. Dhillon, R. Singla, N. S. Rekhi, and R. Jha, “Eog and emg based virtual keyboard: A brain-computer interface,” in Computer Science and Information Technology, 2009. ICCSIT 2009. 2nd IEEE International Conference on. IEEE, 2009, pp. 259–262. [8] M. Vidal, A. Bulling, and H. Gellersen, “Analysing eog signal features for the discrimination of eye movements with wearable devices,” in Proceedings of the 1st international workshop on pervasive eye tracking & mobile eye-based interaction. ACM, 2011, pp. 15–20. [9] A. Dasgupta, S. Chakraborty, P. Mondal, and A. Routray, “Identification of eye saccadic signatures in electrooculography data using timeseries motifs,” in India Conference (INDICON), 2016 IEEE Annual. IEEE, 2016, pp. 1–5.
[10] A. Dasgupta, S. Chakraborty, and A. Routray, “A two-stage framework for denoising electrooculography signals,” Biomedical Signal Processing and Control, vol. 31, pp. 231–237, 2017. [11] H. Prance, Sensor developments for electrophysiological monitoring in healthcare. INTECH Open Access Publisher, 2011. [12] H. Collewijn, C. J. Erkelens, and R. Steinman, “Binocular coordination of human horizontal saccadic eye movements.” The Journal of Physiology, vol. 404, p. 157, 1988. [13] O. Edfors, M. Sandell, J.-J. Van de Beek, S. K. Wilson, and P. O. Borjesson, “Ofdm channel estimation by singular value decomposition,” IEEE Transactions on communications, vol. 46, no. 7, pp. 931–939, 1998.
