TITLE of Invention: IoT based Cortic Pad for sleep analysis.
1. BACKGROUND OF THE INVENTION
Sleep is one of the most integral and unavoidable parts of a human life cycle. An adult human being goes through five stages of sleep, namely, NREM1 (Non-Rapid Eye Movement), NREM2, NREM3, NREM4 and REM (Rapid Eye movement). NREM1 and NREM2 are light sleep stages whereas NREM3 and NREM4 are deep sleep stages. REM is that sleep stage during which the subject starts dreaming. On an average this cycle repeats every 90 minutes while the complete sleep span. During each sleep stage various parameters of human body such as heart rate, brain waves, muscular movements, show behavioural discrepancies. Out of these five stages of sleep there is only one short time span which is apt for waking up the subject. If the subject wakes up in NREM3, NREM4, or REM stages, he/she feels groggy. In other words, the subject may feel dizzy, weak, and unsteady along with a sensation of incomplete sleep and an urge to hit the snooze button. But if woken up at the right moment, the subject's vitals replenish and all those ill-effects can be eliminated. This moment arrives right after the completion of REM stage and the beginning of NREM1, which is light sleep stage.
1.1. EEG AND ELECTRODE PLACEMENT
The Electroencephalogram (EEG) signal is produced as a result of the electrical activity caused by approximately 100 billion neurons of an adult human brain. It has been proven that number of neurons forming groups with similar functions when fire in the same location on the cerebral cortex generates the EEG on the scalp.
If optimal sites are calculated and selected for recording of EEG activity then the probability of success of the Brain Computer Interface (BCI) system increases. Most of the presently used BCI systems rely on non-invasive techniques, i.e., the electrodes are mounted externally on the scalp. This is accomplished by wearing an electrode cap which helps to obtain the EEG signals from the scalp. This approach is largely accepted, as it does not require electrodes to be surgically implanted inside the brain by invasive methods and thus makes BCI practically harmless and user friendly.
1.2. EEG RHYTHMS
Small electrical activities are continuously generated by the human brain in the form of voltage signals. These signals generally in the frequency range from 0.4 Hz to 100 Hz. But usually the fundamental signals range from 0.4 Hz to about 50 Hz. This range is further divided into various bands of frequencies which correspond to different sleep stages. This classification into bands arises because it has been found that all of these bands have their own characteristic significance as explained by American Sleep Association. Deep sleep is associated with the frequency range of 0.5 to 4 Hz which corresponds to the delta band. These rhythms are the slowest waves but have maximum amplitude. With these low but dominant frequencies, the responsiveness to stimuli decreases significantly. The range of 4 to 7 Hz corresponds to theta band, which is dominant in children and in adult during rest or sleep. It is also dominant during events of drowsiness and also in the events of indulging in meditation or in a relaxed or creative state. The band ranging from 8 to 13 Hz is commonly observed in frontal and occipital regions of brain, which corresponds to alpha band. A healthy human brain always shows a predominance of the alpha rhythm when eyes of the subject are closed.

Opening the eyes leads to blocking of the alpha rhythm, while the other bands become more obvious. The EEG waves and their relevance are shown in Table 2.
While recording the EEG signal some artifacts are also associated with it. Artifacts in this context unwanted information signals detected along with the EEG but originates from non-cerebral origin.
Table 1: EEG waves and their relevance.

Brain Rhythm Frequency Range Amplitude Comments
Delta 0.5 - 4 Hz <100uV Deep Sleep Stage, NREM3 and NREM4
Theta 4-7Hz <100uV Light sleep Stages, NREM 1 and NREM2
Alpha 8-13 Hz 20-60uV Drowsy, NREM1 stage
Beta 13-30 Hz <20uV Random and fast activity, awake stage
Sleep Spindles 2 - 14 Hz <30uV Short bursts of 12-14Hz
K-complexes <100uV Very brief large spike activity
Gamma >30Hz <2uV Subject is paying attention
1.3. THE 10-20 ELECTRODE PLACEMENT SYSTEM
The conical neurons under the cortical surface generate varying electrical fields on the surface of the skull. It can be recorded with appropriate electrodes. The combine activity of the neurons called EEG and transferred onto the area of the recording electrode. The potential changes on the cortical surface largely depend on the postsynaptic potentials at the dendrites of the pyramidal neurons. The pyramidal neurons are situated at right angles to the cortical surface thus their impacts are much greater than other neurons. In addition of it, these pyramidal neurons are all orientated in parallel to one another, so the equi-directional potential changes of the neighbouring pyramidal neurons are added. When several pyramidal neurons are simultaneously excited, an EEG deflection can occur. This signal represents a very small potential of approximately 10 jaV to 100 |iV when measured on the scalp and 1-2mV when measured on the surface of the brain.
To access this potential, electrodes mounted on the scalp of an individual in accordance with the 10/20 electrode placement system. This is based on the relationship between the location of the electrode and the underlying area of the cerebral cortex. Three anatomical reference points should be determined before the electrodes are mounted on the skull of the user. These reference points are the positions of the nasion, inion and the preauricular region. The nasion reference point is below the forehead and at the onset of the nose on the skull. The inion reference pointy is a small bony protrusion between the neck and the skull. And the preauricular is located before the cartilaginous protrusion of the auditory canal.

1.4. ELECTRODE CONFIGURATIONS FOR MEASUREMENT
For recording EEG signals, there are two modes unipolar or bipolar in which electrode can be connected. In unipolar mode, either of the electrodes is connected to the positive input of the amplifier and another electrode is attached to a common or reference position. Thus, in unipolar mode, all the negative electrodes are connected to the same common or reference electrode.
In the bipolar configuration, both the electrodes are connected to specific locations over the scalp. The potential difference between a pair of electrodes is measured. A single action potential or neuronal response cannot be registered at the surface of the scalp, and hence any change in potential is measured by the EEG recording is the effect of thousands of neurons firing simultaneously.
2. DETAILED DESCRIPTION OF INVENTION
The overall Op-Amp based system architecture is designed to act as an instrument which can be used for both IoT applications and neuro-research or sleep study all in a small form factor of a power efficient wearable.
2.1. ANALOG FRONT END
The EEG signals picked up by the EEG electrodes are generally in range of microvolts. They need to be amplified in order to make them displayable and distinguishable. The amplifier pick up the weak biological signals and increase their amplitude for further processing. Electrostatic shielding or grounding is used to reduce electric and magnetic interference. Three different configurations of EEG analog front end were designed and tested.
The quality of the signal varies for all these three configurations. The appearance of actual EEG signal is more like random noise. Therefore, while testing these circuit configurations ECG (Electrocardiogram) has been recorded to check for the high frequency noise content of the signal, as the same circuits can be used to tap ECG off human body. The first configuration, i.e., EEG Analog Front End Revision 1, as shown in figure 1, was designed in order to filter out 50 Hz power line noise component from the acquired EEG signal. INA128 was used as the instrumentation amplifier to measure the differential signal tapped off points F7 and F8 near the frontal region of the brain as per the 10-20 Electrode placement system.
A 2-stage twin-t Notch filter with 50 Hz as cut-off frequency was designed using OPA2340 dual op-amps. The second configuration as shown in figure 2, was a full-fledged combination of pre-amplifiers, band pass filter and Driven right leg (DRL) circuit. The DRL circuit is used for inverting and cancelling out the common mode noise signals, including the power line noise induced in the body. This configuration worked on dual power supply +-3V and therefore required 2 batteries, which made the system bulky.
The third configuration as shown in figure 3, did not require a dual power supply, instead it works on a virtual ground. It requires a single battery. The instrumentation amplifier part is same as before but it is followed by a 0.31 Hz 2nd order High Pass Filter and a 8th order Switched capacitor 55.8 Hz Low pass filter. Both of these HPF and LPF work as a band pass filter to filter out low frequency DC noise and High frequency noise respectively.

2.2. DATA ACQUISITION
The EEG signals from the scalp are continuous analog signals. Analog to Digital converter (ADC) is used for digital representation. These signals are pre-processed and then send to a computer which processes them and uses them in order to accomplish the desired tasks. Various filter are used to filter out the power line interference and movement artifacts, which are potentials generated by blinks, muscle movement and cardiac activity. The EEG signals of interest are typically between 2 and 60 Hz.
The revision 3 of front end uses atmega328 and its internal 10-bit resolution ADC for sampling the filtered EEG wave. In the ADC clock settings, the pre-scalar used is 128 which make the clock for the ADC clock to be
16 MHz (main Clock)
ADC clock = \— = 125 KHz
13
Now as each ADC conversion in AVR takes 13 ADC clock cycles, maximum possible sampling rate that can be achieved using AVR chip Atmega328 is
, N 125 KHz Sampling rate (fs) = ——— = 9.615 KHz
Now this acquired data is send over the UART bridge in two ways. It can be sent over the UART-USB bridge to PC for processing the signal in MATLAB as well as it can be sent over BLE to Android device for further processing. Therefore, we also have to use the Serial commands in the Data acquisition chip. Therefore, final effective sampling rate from the processing end perspective becomes fs = 2.341 KHz at a serial baud rate of 115200. Maximum relevant frequency component in the filtered EEG waveform is of 60Hz.
Therefore fm = 60Hz.
Sampling rate 39.01 samples/Hz,
In the final SMD version of the wearable the current AVR chip may get replaced by a faster and 12-bit resolution MSP430. If required, the internal ADC of these chips will not be used, instead an external 24-bit or 18-bit resolution ADC will be used which will communicate on SPI bus with the acquisition chip.
2.3. PCB LAYOUTS AND EXPERIMENTAL SETUP
After 3 revisions stable front end design has been achieved. Figure 4 shows the layout of the front end rev3 PCB, figure 5 shows the PCB trace for the same. Figure 6 shows the final EEG/ECG recording setup. The whole Op-Amp based system architecture is modified to enhance the performance of the system as well as to add in more features to the front-end. Almost most of the front-end filters and amplifiers are replaced by a single chip (ADS 1294) by Texas Instruments, USA. The complete system block diagram is shown in figure 7.
Acquisition Front-End
The acquisition front-end now comprises of a bio-potential ADC chip known as ADS 1294. It
has internal input differential amplifiers internally. It also has sophisticated bandpass filters
internally. It offers 24-bit resolution on 4 channels simultaneously. It even has right leg driver

internally for noise cancellation. The input line of internal RLD circuit is selectable between all channels separately. The chip communicates to the on-board microcontroller via Serial Peripheral Interface (SPI). It has two modes to operate: Continuous Read Data Mode (Command: RDATAC), Normal Read Data Mode (Command: RDATA). The data sequence is 216 bytes long. The first 24-bits are for GPIO states and LEAD-OFF state, rest 192 bits are for 24-bit data of each of the 8 channels. In our case as ADS 1294 is used we only have 4 channels. Therefore, significant data is available in first 120 bits out of 216 bits of data sequence. ADS 1294 also has internal Programmable Gain Blocks with lx, 2x5 3x, 4x, 6x, 8x, 12x selectable gain values. The sampling rate is also programmable which ranges from 200 SPS to 32 KSPS.
Power Management Hub
All the regulators used for regulating different bus voltages are low drop out regulators strategically selected for keeping the system conducive to low voltage batteries such as 3.7 V Li-Ion Battery. For regulating 3.3 V for all the digital ICs on-board, REG103 IC is used. +2.5V and -2.5V needed for analog biasing of ADS 1294. These voltages are regulated by TLV70025 and TPS72325 respectively. And negative voltage is generated by LM2664 (Inverter-charge pump).
Battery Charge Controller
TP4056 is used as the on-board charge controller for lithium ion battery. The charging current is programmable and the charging time also varies accordingly. The value of external resistor Rprog knowing the desired charging current value is calculated by:

The 2 status LEDs (CHRG and FULL) indicate the statuses of the battery and its charge condition. CHRG LED is on when the battery is charging while FULL LED is off in this condition. The charging stops when the charging current reaches 1/10th of the programmed charging current: CHRG LED turns off and FULL LED starts glowing until charge drops below under-voltage threshold again.
Features offered by the Final System
• Has 4 Bio-Signal Channels with 24-bit resolution each.
• Supports Right Leg Drive (RLD) for Common Mode Noise Cancellation.
• Supports sampling rate ranging from 200SPS to 32KSPS.
• Can be used for EEG (Electroencephalogram) and ECG (Electrocardiogram) acquisition.
• Can be easily programmed via USB.
• Can be powered up via USB as well as works on 3.7 V Li-ion Battery suited for wearables.
• Has On-board Battery Charge control.
• Has a very small factor (75mm x 40mm) thereby allowing the user to directly use it as a wearable.
• Has ARM cortex M0+ at its core, with 48MHz clock speed.
• Arduino Compatible (Arduino Zero Profile).
• 2 Serial Ports are available. Serial 0 available for the user to connect external peripherals.
• Has On-Board Bluetooth - BLE 4.0 (also Programmable) connected on Serial 1 port.
• Has On-board 10-bit DAC and 12-bit ADC (multiplexed on same pin).

• 4 LEDs: BLE connectivity, Battery Charging status, Battery Full status and one USER led(user programmable/ used for Debugging).
• Data acquisition can take place via USB as well as can be sent to Phones/Tablets (for Android/IOS apps) directly via Bluetooth.
The Cortic Alarm
The android application named, Cortic Alarm, is the frontend of the proposed system. It will be the HMI (Human Machine Interface) of the whole system. Therefore, it is very necessary to design this application to be as intuitive and user-friendly as possible. The features of the App are as follows:
• Real-Time Monitoring
The app provides real-time EEG data on screen as shown in figure 8. The displayed waveform can be stopped to view important features at any time instance. The data acquired from the BLE bridge is reconstructed on a graph and displayed on the screen. This data is also analysed for feature extraction and the resultant dominating rhythm component of the signal will be notified on the screen under "status". The status will also include corresponding user condition (awake, light sleep, deep sleep) associated with the dominating rhythm component.
• The Adaptive Waking Algorithm
The alarm activity allows the user to perform the settings for alarm assertion time. He/she can change and manage different assertion time profiles. Now feature that sets Cortic Alarm apart from other Alarm apps is that it has the ability to wake the user at the right time of his/her sleep stage. An adaptive waking algorithm is designed to select the appropriate wake time according to the user's sleep stage. Figure 9 gives the flow chart of Adaptive Waking algorithm. The user is allowed to select an Adaptive band while setting a wake time. Example if the user sets the wake time as 5:00 AM and a Adaptive Band of ±15 Minutes, then the app will wake up the user at the moment he/she goes from REM to NREM1 stage between 4:45 AM and 5:15 AM. If this sleep stage transition does not occur in this adaptive band then the app will assert a fail-safe alarm at 5:16 AM, irrespective of the user's current sleep stage. Whilst asserting the alarm as per the REM to NREM1 transition in adaptive band, the algorithm also saves the time points in user profile database (as shown in figure 10) to gradually develop a high probabilistic time point and suggest the user the next time he/she sets another alarm.
• The Sleep Timeline
The sleep timeline Activity gives a brief visualization of the sleep stage behaviour derived by the user whilst the previous sleep session. The y axis shows the Stages, REM, NREM1, NREM2, NREM2, NREM3, NREM4 and Awake stage, while the X axis is the time axis.
3. RESULTS
The system is designed for acquiring EEG, but the same configurations can be easily used for ECG acquisition too. Raw EEG is very much similar to noise when observed on screen. Therefore, to validate the functionality of the acquisition system we must observe the noise-free is the ECG acquisition. Therefore, the figure 11 shows the ECG wave data acquired from revision 3 of the op-amp based system with and without right leg drive. There is a significant improvement in the signal quality when RLD is used. The figure 12 shows the analysis of channel 1 EEG data using MATLAB. The sampling rate is 500 SPS. The data is received serially-with a baud-rate of -115200. The data received is corrupted with power line noise but

is filtered using an FIR filter on MATLAB. The FFT clearly shows a peak at 25 Hz, which indicated the dominance of Beta Waves.
REFERENCES:
1. American Sleep Association - Sleep Info: http://www.sleepassociation.org/index.php? p=patients , Sleep Stages and Sleep Disorders, sleep hygiene, Importance of Sleep, Dreaming and REM Sleep, Sleep Products.
2. Vaibhav Gandhi , "BrainComputer Interfacing for Assistive Robotics", Elsevier, Science and Technology Books, Academic Press, 2015.
3. Surya Darma Ridwan, Robert Thompson, Budi Thomas Jap, Sara Lai, Peter Fischer, "Single Channel Wireless EEG: Proposed Application in Train Drivers ", African Journal of Information and Communication Technology, Vol. 5, No. 2, June 2009.
4. F. Sharbrough, G. Chatrian, R. Lesser, H. Luders, M. Nuwer and T. Picton, "American Electroencephalographic Society guidelines for standard electrode position nomenclature," J. Clin. Neurophysiol., vol. 8, pp. 200-202, 2011.
5. Chen, Xun and J. Z. Wang, "Design and implementation of a wearable, wireless EEG
recording system", "Bioinformatics and Biomedical Engineering,(iCBBE) 2011 5th International Conference," in IEEE, 2011.
6. Ying-Wen Bai, Wen-Yang Chu, Chien-Yu Chen, Yi-Ting Lee, Yi-Ching Tsai and Cheng-Hung Tsai , "Adjustable 60Hz Noise Reduction by a Notch Filter for ECG Signals ", Instrumentation and Measurement Technology Conference,Como. Italy. 18-20 May 2004.
7. Sapana M Adhalli, H Umadevi, Guruprasad S P and Rajeshwari Hegde, "Design and Implementation of EEG Signals Analysis on FPGA", International Journal Of Engineering And Computer Science, Vol. 5, No. 5, pp. 16658-16667, May 2016.
8. K. Amarasinghe, D.Wijayasekara, M.Manic , "EEG Based Brain Activity Monitoring using Artificial Neural Networks" IEEE International Conference on Human System Interactions, pp. 61 - 66, 2014.
9. Veerendra Dasari, "EEG Acquisition System on Mobile Platform",Master's Theses, Western Michigan University, Paper 118, 2013.
10. Jonathan Garza, Yuezhe Li, Yuchou Chang, Hong Lin, "A Real Time EEG Analysis System", IEEE International Conference on Data Science in Cyberspace (DSC), 02 March 2017.

4. SUMMARY OF THE INVENTION:
The Proposed sophisticated acquisition system has been designed after several revisions and after trying out several other architectures. The final system possesses number of features as compared to its predecessors. EEG signal can now be obtained from 4 channels. The acquired electroencephalogram readings are used for further analysis. The analysis part of the project needs further developments to be made.
The android application is also developed to work in sync with the hardware system. The connectivity with the on-board BLE module with the android application is developed with modifications in the BLE API. These accomplishments enable the IoT feature of the invention.

CLAIMS:
We claim that:
1) the proposed system can be used for EEG (Electroencephalogram) and ECG (Electrocardiogram) acquisition.
2) the proposed system is an Internet of Things (IoT) enabled wearable brain computer interface design.
3) the proposed system executes an adaptive waking algorithm which can wake-up the user at the right moment of his/her sleep stages.
4) the proposed system can be powered up via USB as well as works on 3.7V Li-ion Battery suited for wearables.
5) the proposed system has On-board Battery Charge control.
6) the proposed system provides data acquisition via USB. The data can be as well sent to Phones/Tablets (for Android/IOS apps) directly via Bluetooth.
7) the proposed system has the app to provides real-time EEG data on screen and its analysis.