📄 This portfolio is a self-contained HTML document with all images embedded.
Contains 10 images extracted from PDF.
Videos require Google Drive access to play.
Core Technology & Measurements
The proposed device is a pair of lightweight smart glasses that combine three sensing modalities: dry frontal–temporal EEG, compact optical fNIRS over the prefrontal cortex, and integrated eye-tracking cameras in the frame. This hybrid design leverages the complementary strengths of each technology for everyday cognitive state monitoring.
Sensing modalities and placement. Dry EEG electrodes are embedded along the inner temples and upper frame, approximating positions over lateral prefrontal and temporo-parietal regions (similar to AF7/AF8/TP9/TP10 in the 10–20 system). Consumer headsets such as Muse use a similar frontal montage and have shown acceptable quality for event-related potentials and brain–performance indices in ambulatory settings (Byrom et al., 2018; Krigolson et al., 2017). fNIRS optodes (sources and detectors) are placed along the upper rim of the glasses, targeting dorsolateral prefrontal cortex (DLPFC), a region strongly implicated in cognitive control and workload. fNIRS has been highlighted as a practical, portable brain-monitoring tool in digital neurology and in multimodal systems that combine NIRS and EEG to assess neurovascular coupling (Phillips et al., 2023; Pinti et al., 2021; Li et al., 2022).
Infrared eye-tracking cameras are embedded near the nose bridge and in the lower frame, capturing binocular gaze, saccades, fixations, blinks, and pupil size. Normative data show that eye-movement metrics such as saccade latency and fixation stability are reliable markers of cognitive and neurological status when collected with head-mounted systems (Kullmann et al., 2021).
Figure 1. AI-generated prototype image of the proposed glasses (ChatGPT 5.1).
Page 1 visual contentPage 2 visual content
System Process Diagram
Figure 2. AI-generated system process diagram (ChatGPT 5.1). A practical configuration is 8 EEG channels, 8–10 fNIRS channels (4 sources × 4–5 detectors), and two inward-facing cameras for eye tracking. This setup should offer a good compromise between spatial coverage, comfort, and power consumption for consumer use.
User-facing metrics. From these raw signals, the glasses would report a small set of interpretable metrics:
Focus / attention index: based on frontal midline theta, beta power, and gaze stability during task engagement. Cognitive state decoding using wearable EEG and peripheral signals has already been demonstrated in n-back paradigms (Azgomi et al., 2023) and in wearable workload monitoring (Zanetti et al., 2022).
Cognitive load & mental fatigue: derived from changes in prefrontal fNIRS oxy-Hb/deoxy-Hb, frontal theta/alpha ratios, and increasing blink rate or pupil dilation during sustained tasks. Wearable fNIRS platforms have shown reliable individual-level functional patterns suitable for precision mental health applications (Phillips et al., 2023; Victorio et al., 2025).
Stress / relaxation index: combining EEG alpha/beta ratios, autonomic proxies (e.g., blink rate, pupil diameter) and behavioural markers (restlessness inferred from micro-saccades). Similar multimodal wearable studies using EEG and peripheral signals have successfully inferred arousal and performance states in everyday “safe actuation” interventions (Azgomi et al., 2023).
Page 3 visual contentPage 4 visual content
Hardware Design & Form Factor
Physical design. The glasses resemble slightly thicker premium frames rather than lab equipment. The front frame uses lightweight TR-90 nylon or titanium; temples integrate flexible printed circuit boards, batteries, and EEG contact points. Nose pads and ear-contact surfaces are medical-grade silicone. The weight target is 40–50 g, comparable to camera-equipped smart glasses. Adjustable nose pads and spring-loaded hinges support a wide range of head sizes for all-day wear.
Dry EEG electrodes are gold- or Ag/AgCl-coated spring contacts, gently pressing on skin at the temples and behind the ears; foam or comb-like geometries improve contact without gels and have been widely used in wearable devices (Byrom et al., 2018). fNIRS LEDs and photodiodes are encapsulated in soft, opaque elastomer “pods” along the upper frame, with shallow contact to the forehead, consistent with current wearable fNIRS biosensors (Victorio et al., 2025).
Technical specifications. A feasible first-generation target is:
Battery life: 8–10 hours of intermittent sensing (e.g., 30–50% duty cycle) using a ~300–400 mAh Li-polymer cell distributed in both temples.
Connectivity: Bluetooth Low Energy for continuous streaming to a smartphone; optional Wi-Fi only when on a charging dock for firmware updates and secure cloud sync.
Processing: a low-power MCU and DSP handle filtering, feature extraction, and lightweight classification; more complex models run on the phone.
Page 5 visual contentPage 6 visual content
User Experience & Software
Smartphone app and feedback. The companion app presents three levels of information:
Glanceable state: a simple dial or “traffic light” showing current focus/overload/under-stimulated states, plus a small trend arrow.
Session view: during “work sessions,” users can see time-series graphs of focus, load, and stress, as well as eye-movement metrics (e.g., mind-wandering episodes indicated by off-screen gaze and frequent blinks).
Daily/weekly summaries: aggregate statistics (hours in productive focus, time spent overloaded, recovery quality) and contextual insights — e.g., “You sustain high focus for ~35 minutes before overload; your best deep-work window is 9–11 AM.”
Feedback is not constant; instead, micro-interventions are triggered when metrics cross thresholds for several minutes (e.g., intense overload with rising stress), similar in spirit to safe everyday interventions that modulate cognitive states via music or coffee (Azgomi et al., 2023).
Use cases and “killer app”.
Productivity optimization: real-time and retrospective insights about when to schedule deep work, when to take breaks, and which tasks are most cognitively draining.
Mindfulness and stress regulation: EEG/eye-based biofeedback during breathing or meditation exercises, augmented by fNIRS prefrontal activity as an additional marker of engagement.
Early cognitive-change flags (opt-in): periodic 5–10 minute gamified tasks probing working memory and oddball P300 responses, inspired by portable EEG MCI work (Smith, 2022) and remote digital screening that outperforms traditional tools for detecting cerebral amyloid status (Thompson et al., 2023).
Future clinical integrations: longer-term, the same hardware could support remote dementia-risk monitoring or depression treatment response in partnership with emerging wearable brain-monitoring technologies, such as fNIRS-based dementia screening devices (Optics.org, 2023; Victorio et al., 2025).
Page 7 visual contentPage 8 visual content
Validation & Scientific Credibility
To avoid the “junk science” reputation of some consumer brain gadgets, validation must be central.
Study design.
Lab validation: A cohort of participants performs standardized tasks (eyes-open/closed, n-back, sustained attention, relaxation periods) wearing both the glasses and a research-grade EEG and fNIRS cap. Agreement between systems is assessed via spectral features, ERPs, and hemodynamic responses, leveraging analytical frameworks for combined EEG–NIRS (Pinti et al., 2021; Li et al., 2022).
Cognitive-state prediction: Using protocols similar to prior wearable workload studies (Zanetti et al., 2022; Azgomi et al., 2023), models predict task difficulty, performance (accuracy and reaction time), and subjective workload. Pre-registered analyses and rigorous cross-validation guard against overfitting.
Field trials: Knowledge workers wear the glasses for several weeks; metrics are correlated with ecological outcomes (self-reported fatigue, productivity logs, error rates) and digital performance tests aligned with emerging next-generation cognitive assessment frameworks (Hall, 2023; Thompson et al., 2023).
Regulatory and communication strategy. Initial positioning is as a wellness and productivity device with no diagnostic claims, following the model of existing consumer EEG products and mobile brain-monitoring tools (Byrom et al., 2018; Krigolson et al., 2017). If future clinical trials show robust sensitivity to MCI or dementia trajectories — paralleling portable EEG P300 and remote digital screening results (Smith, 2022; Thompson et al., 2023) — medical-device clearance for specific indications could be sought. For consumers, the app must clearly state that metrics are probabilistic and trend-based, not definitive diagnoses, and offer confidence ranges plus education on limitations (e.g., “High stress index may be influenced by motion or poor contact today”).
Page 9 visual contentPage 10 visual content
Business & Market Reality
Cost and price point. Using commodity components (CMOS cameras, BLE chipsets, Li-poly batteries, standard optics) with custom EEG/fNIRS front-ends, early bill-of-materials (BOM) might be around USD 120–150 at moderate volume, suggesting a v1 retail price near USD 350–400. With scale, simplification of the optical array, and multiple SKUs (e.g., EEG + eye only vs. full EEG+fNIRS), the platform could approach the < USD 200 price band typical of advanced smartwatches, especially as wearable optical and EEG chips continue to commoditize (Byrom et al., 2018; Victorio et al., 2025).
Market positioning and competitors. Primary targets are knowledge workers and high-stress professionals (software engineers, managers, traders, clinicians) interested in productivity, burnout prevention, and long-term brain health. Competing devices include EEG headbands such as Muse, wearable EEG headsets from Emotiv and Neurosity, and research-grade EEG-eyewear prototypes (Byrom et al., 2018; Vourvopoulos et al., 2019; Kosmyna et al., 2019). fNIRS-based wearables and home brain-monitoring platforms are just emerging (Phillips et al., 2023; Victorio et al., 2025; Optics.org, 2023).
The proposed glasses differentiate themselves by:
Seamlessly fitting into existing eyewear habits.
Combining EEG + fNIRS + eye tracking in a single everyday form factor.
Focusing on actionable daily coaching rather than abstract “brain scores.”
Privacy and ethics. Continuous brain and gaze monitoring raises serious ethical concerns, especially around employer access and mental-state surveillance. Following best-practice recommendations for digital brain data in clinical and mental-health contexts (Byrom et al., 2018; Thompson et al., 2023), the design assumes:
All raw and derived brain data are encrypted at rest and in transit; raw data stay on device or user-controlled storage by default.
No third-party sharing (including employers or insurers) without explicit, revocable consent, and no real-time dashboards for managers.
On-device or on-phone processing wherever possible, with optional anonymized cloud analytics.
Clear, human-readable privacy policies and in-app explanations of what is and is not inferred help prevent misuse and build trust. Long-term success in this space will depend as much on ethical stewardship of neural data as on technical performance.
📚 References
Azgomi, H. F., Branco, L. R. F., Amin, M. R., Khazaei, S., & Faghih, R. T. (2023). Regulation of brain cognitive states through auditory, gustatory, and olfactory stimulation with wearable monitoring. Scientific Reports, 13 (1), 12399. https://doi.org/10.1038/s41598-023-37829-z
Byrom, B., McCarthy, M., Schueler, P., & Muehlhausen, W. (2018). Brain monitoring devices in neuroscience clinical research: The potential of remote monitoring using sensors, wearables and mobile devices. Clinical Pharmacology & Therapeutics, 103 (6), 942–948. https://pmc.ncbi.nlm.nih.gov/articles/PMC6032823/
Kosmyna, N., Morris, C., Nguyen, T., Zepf, S., Hernandez, J., & Maes, P. (2019). AttentivU: Designing EEG and EOG Compatible Glasses for Physiological Sensing and Feedback in the Car. Proceedings of the 11th International Conference on Automotive User Interfaces and Interactive Vehicular Applications, 355–368. https://doi.org/10.1145/3342197.3344516
Krigolson, O. E., Williams, C. C., Norton, A., Hassall, C. D., & Colino, F. L. (2017). Choosing MUSE: Validation of a low-cost, portable EEG system for ERP research. Frontiers in Neuroscience, 11, 109. https://doi.org/10.3389/fnins.2017.00109
Kullmann, A., Ashmore, R. C., Braverman, A., Mazur, C., Snapp, H., Williams, E., Szczupak, M., Murphy, S., Marshall, K., Crawford, J., Balaban, C. D., Hoffer, M., & Kiderman, A. (2021). Portable eye-tracking as a reliable assessment of oculomotor, cognitive and reaction time function: Normative data for 18–45 year old. PloS One, 16 (11), e0260351. https://doi.org/10.1371/journal.pone.0260351
Li, R., Yang, D., Fang, F., Hong, K.-S., Reiss, A. L., & Zhang, Y. (2022). Concurrent fNIRS and EEG for Brain Function Investigation: A Systematic, Methodology-Focused Review. Sensors (Basel, Switzerland), 22 (15), 5865. https://doi.org/10.3390/s22155865
Optics.org. (2023). Wearable brain monitor could spot dementia risk earlier. Retrieved November 16, 2025, from https://optics.org/news/16/9/33
Phillips V, Z., Canoy, R. J., Paik, S., Lee, S. H., & Kim, B.-M. (2023). Functional Near-Infrared Spectroscopy as a Personalized Digital Healthcare Tool for Brain Monitoring. Journal of Clinical Neurology (Seoul, Korea), 19 (2), 115–124. https://doi.org/10.3988/jcn.2022.0406
Pinti, P., Siddiqui, M. F., Levy, A. D., Jones, E. J. H., & Tachtsidis, I. (2021). An analysis framework for the integration of broadband NIRS and EEG to assess neurovascular and neurometabolic coupling. Scientific Reports, 11, 3977. https://doi.org/10.1038/s41598-021-83420-9
Smith, H. H. (2022). Assessing mild cognitive impairment using portable electroencephalography: The P300 component. The Arbutus Review, 13 (1). https://doi.org/10.18357/tar131202220753
Thompson, L. I., Kunicki, Z. J., Emrani, S., Strenger, J., De Vito, A. N., Britton, K. J., Dion, C., Harrington, K. D., Roque, N., Salloway, S., Sliwinski, M. J., Correia, S., & Jones, R. N. (2023). Remote and in-clinic digital cognitive screening tools outperform the MoCA to distinguish cerebral amyloid status among cognitively healthy older adults. Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring, 15 (4), e12500. https://doi.org/10.1002/dad2.12500
Victorio, M., Dieffenderfer, J., Songkakul, T., Willeke, J., Bozkurt, A., & Pozdin, V. A. (2025). Wearable Wireless Functional Near-Infrared Spectroscopy System for Cognitive Activity Monitoring. Biosensors (Basel), 15 (2), 92. https://doi.org/10.3390/bios15020092
Vourvopoulos, A., Niforatos, E., & Giannakos, M. (2019). EEGlass: An EEG-eyeware prototype for ubiquitous brain-computer interaction. In UbiComp/ISWC ’19 Adjunct: Adjunct Proceedings of the 2019 ACM International Joint Conference on Pervasive and Ubiquitous Computing and the 2019 International Symposium on Wearable Computers (pp. 1–6). https://doi.org/10.1145/3341162.3348383
Zanetti, R., Arza, A., Aminifar, A., & Atienza, D. (2022). Real-time EEG-based cognitive workload monitoring on wearable devices. IEEE Transactions on Biomedical Engineering, 69 (1), 265–277. https://doi.org/10.1109/TBME.2021.3092206
Other Course Work
📊 Assignment 1: EEG Analysis
📓 PSYCH403_assignment1_eeg_filtering_Yan - Yan Luo.ipynb
Jupyter Notebook
Interactive Python notebook with code, visualizations, and analysis
Concept E xplanation As a student, I personally relate to how difficult and how important it is to pay attention and sustain my attention on something. However, o ften, we do not receive the reward s immediately after giving the attention where it is required. Attention can be very exhausting and boring depending on the tasks at hand. This attention - based snake game was an attempt to make the process of paying attention fun, interesting, and interactive. After all, who does not want to be entertained with games while training on probably one of our most precious finite resources ? T echnical I mplementation With the help of Cha t GPT, the final code incorporates the code from the template Neural Mandala and the code from the snake game example provided on p5.js. 1.At the global level The code starts with defining the angle of the rotation of the mandala. The snake variables define the grid that the snake moves ( 30 x 30 units) , the starting segments of the snake (10 units), and the x and y coordinates of the location the snake will start. The Fruit variable has to be defined for later processing , and fruitAlpha variable allows the fade - in glow when it appears on the screen. 2.Vector helpers Considering the p5.Vector was not available on BrainImation’s environment ; this section was added to define and calculate the vectors in the grid for locating the snake and fruit position s. 3. Setup() The setup section defines the HSB color mode, and allows the game to start with a snake with a fruit randomly appeared on the screen. 4.Draw() 1) EEG inputs – Obtain the EEG data from the attention and the beta data set. 2) Mandala section – Centre the mandala on scree n with growing layers as there is more focus (minimum is 2 layers and the maximum is 10). T he beta data control s the expansion/contraction of the mandala ; the more beta activity leads to a faster “ bre athing”
effect . The color dynamic of the mandala includes t he hue and brightness of the mandala change when the attention changes , and the size of the mandala changes with the beta. 3) Rotation speed – The speed of mandala rotation is controlled by the level of attention. More attention there is, the faster it will spin. 4) Snake – The snake’s head will move towards the fruit automatically. The speed of the snake also corresponds with the level of attention. A higher level of attention leads to a faster speed of movement of the snake. The snake’s movement will be confined within the grid. 5) Fruit collis i on & Fade - in Fruit – Defines when the snake eats a fruit, the snake grow s, and the score increases by 1. The fruit will appear with a fade - in effect at a random location on the screen. 6) Snake rendering – T he color and the glow intensity of the snake will increase as there is more attention. 7) Fruit rendering – D raw the fruit as an orange ball (with a defined size to aslign with the size of the snake ) on the screen. 8) Score – A higher score indicates sustained attention and offers instant rewards to reinforce the higher attention behaviors. C hallenges F aced The first and most difficult one was to choose and decide what I want ed to do for this assignment. After exploring both the suggested templates and examples on p5.js for a while, I finally decided to make an interactive game that is exciting to be engaged in the process. The next challenge was to make the code work in BrainImation’s live editor environment. There were several times the error s were show n , either because the environment does not use these functions or some of the functions were not defined. W hat W as L earned ChatGPT has been a great reso u rce i n the process . It is not difficult to ask AI to help generate useful code . I, as the designer of the product , still need to have a clear idea of what this product is going to be like and for wha t demographic. The thinking and the trial - and - erro r processes are still essential for developing products. I also needed to understand the function and the purpose of each section of the c o de in order to make further improvements. Additionally, mistakes and warning signs are less intimidating than they seemed before. R ecognizing these could be learning opportunities rather than reading them as discouragement was also reflective for me.
🎥 PSYCH403_Assignment_2_Video_Yan_Luo.mp4
💡 Videos require Google Drive access. Open in new tab if it doesn't load.
PSYCH 403 – Midterm Name: Yan Luo Student ID: 1768861 Part 1 Concept - The BCI concept for this visualization combined “adaptive difficulty system” and “multi - state classification”. Logic - When the alpha level is high, it suggests a calmer and relaxed mental state ; on the other hand, when the theta level is high, it indicates a tense or stressed mental state. Therefore, the alpha/beta ratio was used to define the calmness level, and the theta/beta ratio define d the tension level. Low in either ratio (high beta ) would suggest a focused state. A high alpha level would bring a more radiant, smoother breathing background of the BCI , more raindrops , more and brighter splash particles that radiate further , and the brighter blue s for the ripples ( i.e., colors in aqua, azure and bright blue ) . Conversely, a higher theta/beta r atio will make fewer raindrops , fewer and tighter splashes, and the darker blues for the ripples (i.e., colors in cornflower, steel blue, soft navy). A higher beta (focused state) will add a brighter shimmer to ripple edges. What makes i t different - Instead of just using the brain wave data to update the visualization, the interactive nature of this BCI can serve as a breathing exercise tool . A s the interface receives live EEG data from the user , it demonstrate s the current mental state ( increasing awareness) , and it encourages the user to create more complex rainfall /storm - like visual behaviors ( acting as positive reinforcement ) . What was learned from implementing it - The order of the events and the realistic representation matter. ChatGPT used straight streaks representing the raindrops, and the raindrops fell after the ripple s appeared. When these “mistakes” occurred in the code, I was often surprised that AI did not consider and capture the sequence of natural events and their realistic or common representation when generating the code. However, depending on the structure or algorithm of ChatGPT, it may
be reasonable to assume it focus ed the calculating resources on what was asked and did not deploy extra resources to guess my intentions and close the logic loop for me . Link to p art 1 video:
Part 2 ERP component - N170 was selected for this ERP experiment. The stimulus manipulation - The schematic face was inverted in half of the trials . The order of the upright and inverted faces was randomized. A veraging system explanation – a) Timing for each trial 0 – 1000 ms - baseline 1000 – 1050 ms - Face stimulus onset (randomly upright or inverted, 50% of probability of each) 1050 – 1200 ms - B lank interval 1200 – 1050 ms - Object stimulus onset 1250 – 1500 ms - Inter - trial interval b) Data collection for each stimulus event The - 200 – 800 ms of EEG data were collected for each stimulus onset for all four channels (YP9, AF7, AF8, TP01) . This duration of 1000 ms is considered one ERP epoch. c) Implementing baseline correction The mean amplitude from - 200 – 0 ms is subtracted for each epoch to ensure all trials start close to 0 μV . d) Sampling rate The data sampling came from the getRawChannel data from the Brain Imation website and each channel ( i.e., YP9, AF7, AF8, TP01) at a 256 Hz sampling rate.
e) Averaging method It uses a cumulative averaging method. Each epoch from the first trial will contribute to the averaging process. Random n oises should be cancel led in the data through the averaging process . Predic tion s with real EEG data – Over at least 20 trials, the average EEG data should show a significant negative depletion around 170 ms after the on s et of the face stimuli , which shows the face perception and the structure encoding of faces. Additionally, since it mainly occurs over the posterior temporal cortex, the TP01 channel may be expected to have stronger signals than other channels . Lastly, considering an inverted face would lead to a delayed and larger N170 , the upright face stimulus would produce a moderate amplitude and occur around 170 ms , and the inverted face stimulus would produce a n even larger negative amplitude , which may appear around 190 - 200 ms due to disrupted face processing . Link to part 2 video:
