WellPD-Inspired Parkinson's Motor Symptom Monitor: Wrist IMU Tremor Tracking and Bradykinesia Scoring

Parkinson's disease changes how a person moves from hour to hour: tremor comes and goes, movements slow down, and involuntary movements appear, often in patterns that a single clinic visit cannot capture. Doctors and researchers therefore value continuous, objective movement records taken at home over days and weeks, and that need is why wrist-worn symptom monitors have become a serious product category. In late October 2026 Empatica launched WellPD, an FDA-cleared and CE-marked wrist device with a companion app that tracks tremor, dyskinesia, bradykinesia, sleep and activity, and summarises them daily, weekly and monthly, with shipping from 23 November 2026. This student project translates that headline honestly to lab scale: a wrist unit that samples motion, scores a tremor index only when the hand is verifiably still, times a finger-tap test as a slowness proxy, and plots both against manually logged dose times, so a student can demonstrate fluctuation over a day without claiming any medical capability.

The end-to-end mechanism runs from wrist motion to a scored, explainable daily record in four guarded stages, each with a stated rule an examiner can test. First, the MPU6050 streams acceleration and rotation to the ESP32 at a 100 Hz design target, and the firmware watches gyro magnitude against a fixed threshold. Second, only when the hand has stayed below that threshold for 2 seconds does the rest gate open, and only then does the firmware compute FFT band power in the 4 to 6 Hz band and update the tremor index, so walking can never raise the resting score. Third, the wearer runs the two guided tasks: a 10-second tap test through the wrist button and a slow guided reach, which together produce the bradykinesia proxy numbers. Fourth, the ESP32 serves the dashboard over Wi-Fi with the live trace, daily summary, dyskinesia proxy flag, overnight log, and dose times together, exporting the day as CSV.

1. The builder assembles the wrist unit: ESP32 board, MPU6050 module, small OLED, tap-test button, and Li-Po cell wired inside a 3D-printed-style enclosure on a wrist strap, and checks the I2C link and the OLED display.
2. The builder calibrates the unit: the wearer sits with the hand still for the rest-gate check, walks a short distance to confirm walking closes the gate, and runs one tap test to confirm the button counts cleanly.
3. For the rest-gated tremor demonstration, the builder holds the hand still until the dashboard rest-gate indicator opens, watches the tremor index update in the 4 to 6 Hz band, then walks to show the gate closing and the score freezing.
4. For the bradykinesia proxy demonstration, the builder runs the 10-second tap test, records taps per 10 seconds, then performs the guided reach task and reads the movement-speed score from the dashboard.
5. For the fluctuation demonstration, the builder logs two or three manual dose times across the session and shows the tremor curve plotted against those markers on the dashboard, describing it strictly as a correlation display.
6. For the overnight proxy demonstration, the builder leaves the unit logging movement counts overnight or across a long still period, then reads the movement-count log and the daytime step count on the dashboard.
7. The builder exports the day's CSV, points to the rest-window count, the tremor index values, the tap-test result, and the dose-time markers, and reports honestly in the report which windows were rest-gated and which segments were excluded as movement.

Tech stack:

• ESP32 · dual-core controller and Wi-Fi dashboard host
• MPU6050 IMU · 6-axis motion sensing for tremor and activity
• FFT band-power scoring · 4 to 6 Hz resting-tremor band analysis
• Rest-gate rule · 2-second still-hand gate before scoring
• OLED wrist display · live tremor index and rest-gate state
• Tap-test button · 10-second tap counting and dose-time logging
• Wi-Fi web dashboard · live trace and daily summary display
• CSV export · daily indices and dose-time markers for the report

This monitor combines six demonstration measurements into one wrist unit and one dashboard, so the review story is a complete chain, not one reading. The wrist unit samples the MPU6050 sensor at a 100 Hz design target and streams the data to the ESP32, which runs the scoring rules. A rest gate guards the tremor score: the score updates only after the gyro magnitude has stayed below a fixed threshold for 2 seconds, so walking and arm swings cannot leak into a resting-tremor number. A tap test gives a second, independent proxy for slowness, counting finger taps in 10 seconds through the wrist button, while a guided reach task adds a movement-speed score from gyro magnitude. The dashboard turns those indices into a visible day: a live tremor trace, a daily summary with a stated demo formula for time with mild or no symptoms, dose-time markers plotted against the tremor curve, and a CSV export for the report.

• [Rest-gated tremor index] The ESP32 samples the MPU6050 at 100 Hz (design target) and computes FFT band power in the 4 to 6 Hz resting-tremor band, scored only while the rest gate is open after 2 seconds of still-hand gyro data, so walking does not contaminate the score — implemented in the base build.
• [Bradykinesia proxy tap test] A wrist button times a 10-second finger-tap test and reports taps per 10 seconds plus a movement-speed score from gyro magnitude during a guided reach task, both labelled demonstration proxies, never clinical bradykinesia measures — implemented in the base build.
• [Dyskinesia proxy flag] A simple involuntary-movement variability indicator during rest windows flags raised movement variability, explicitly labelled a proxy indicator and never a clinical dyskinesia measure — implemented in the base build.
• [Sleep and activity proxy log] Overnight movement counts and daytime step and activity counts from the IMU give an actigraphy-style proxy log, with no sleep staging and no sleep-quality scoring claimed — implemented in the base build.
• [Medication-time log with fluctuation plot] Manually tapped dose times are plotted against the day's tremor index so a student can see fluctuation around dose times, as a correlation display only, never an efficacy claim about any medicine — implemented in the base build.
• [Live dashboard with daily summary] The Wi-Fi web dashboard shows a live tremor trace, a rest-gate indicator, a daily summary with a WellMove-style time-with-mild-or-no-symptoms demo index whose formula is stated on the page, and a CSV export of the day's indices — implemented in the base build.
• [On-wrist OLED and tap button] A small OLED shows the live tremor index and rest-gate state at a glance, and the tap-test button doubles as the dose-time log button, so the unit works standalone before the dashboard is opened — implemented in the base build.
• [Two-wrist comparison] Future scope only, not implemented: wearing a matched unit on each wrist and comparing tremor indices side by side, listed as an extension rather than a claimed feature, since the base build covers one wrist and one wearer per calibration.

• Complete fabricated wrist unit (ESP32 board, MPU6050 module, small OLED, tap-test button, Li-Po cell, 3D-printed-style enclosure, wrist strap)
• Wi-Fi dashboard software (live tremor trace, rest-gate indicator, tap test, dose-time log, daily summary, CSV export)
• Setup and calibration guide (rest-gate check procedure, walking-exclusion check, tap-test procedure, dose-time logging worksheet, CSV reading worksheet)
• Project report PDF (wearable symptom-monitoring background, rest-gate and FFT band design, proxy-index design, demonstration results, honest measurement limits)
• PPT presentation for final review
• Viva Q&A preparation document (resting tremor band choice, rest-gate design, proxy versus clinical measures, WellPD comparison limits, single-wrist limits, fluctuation-plot honesty)