From EEG and fragrance to a testable in-person POC.

I led the cross-modal proposal, core experience, experimental flow, device selection and data-visualisation planning, coordinating with hardware, engineering and brand stakeholders on POC and commercial-delivery proposals.

The work starts with the product problem, then develops the journey, service coordination, records, system responsibilities, candidate refinement and prototype.

Three connected design questions guide this case: how stimulus relates to feedback, how devices and on-site service coordinate, and how users inspect and adjust.

Current outputs are experiment/service proposals, system responsibilities and an interaction prototype. Charts and optimisation logs are demonstrations; actual acquisition, integration and cross-modal associations still need validation.

Notes and ingredients describe a scent, but do not establish why it fits someone. The same scent can evoke different responses. Candidates need context: use setting, liking, intensity and sensitivity information.

Staff guide fitting, pacing and explanations. Research/technical teams need signal-quality checks and correspondence between samples and records.

The product therefore needs to explain how users express feelings, what organises candidates, and how staff communicate the process.

Starting point: fragrance and meditation content

Combining scent with content creates an experience entry point, but still needs an account of individual differences and candidate rationale.

Add EEG as a reference input

EEG opens a direction for researching individual responses, while adding requirements for fitting, quality checks, record alignment and interpretation. It needs to be understood alongside subjective reports.

Move into an in-person space

The environment, staff and terminal support fitting, acquisition, sampling and adjustment, placing hardware and software in an organised workflow.

My design scope expanded from content/screens to experiments, service, hardware/software coordination and delivery conditions.

Can one session produce traceable records?

Questionnaires, scent samples, presentation conditions, acquisition and round-by-round feedback must relate to one session so a report can return to the process.

Are associations between inputs usable?

EEG, GC-MS composition information and subjective feedback need correspondence under consistent conditions before signal quality, individual differences and association stability can be examined.

How does feedback enter the next choice?

A candidate should not be a one-time conclusion. Acceptance, rejection, intensity and directional changes need to affect subsequent choices so users can continue correcting the experience.

1. 01Check in and explain
2. 02Fit and calibrate
3. 03Acquire baseline
4. 04Match candidates
5. 05Experience
6. 06Collect feedback
7. 07Generate report

Each step connects user action, front-stage service, back-stage processing and technical support. The user experiences one journey; the team coordinates samples, device status, records, feedback and delivery.

The service blueprint turns system capabilities into who does what and when. Check status before acquisition, associate sampling with a sample ID, record why an adjustment occurred, and link the report to the actual session.

Baseline

Record resting state and subjective feelings, inspect acquisition quality and establish a starting point for individual comparison.

Fragrance stimulus

Associate sample identity, presentation time and intensity with signal changes, recording liking and comfort.

Recovery

Observe post-stimulus changes alongside subjective feedback. A single physiological change cannot directly establish liking or efficacy.

I organised relationships between physiological information, samples/devices and subjective reports. Time, conditions and adjustments need to remain traceable for later analysis and reporting.

Sensing collects EEG reference and preference input.

Data organises signals, features, samples and GC-MS information.

Decision handles preference constraints, candidate directions and explanatory evidence.

Execution plans diffusion, sound/light, intensity and pacing controls.

Interaction provides acquisition guidance, monitoring, adjustment and reports.

These are proposed responsibilities, not a completed production integration. They support discussion of dependencies: acquisition quality affects usable records; subjective feedback affects candidates; device state and adjustment history need to be understandable in the interface.

Device planning covers portable/research EEG, acquisition/control, diffusion, audio, lighting and PC/tablet terminals for experimental and delivery needs. Actual capabilities and integration depend on implementation.

The proposed recommendation mechanism connects composition/sample information, the target experience, signal references and subjective feedback in four stages:

1. Baseline and initial selection: use available information and preference constraints to propose candidates with reasons.
2. Sampling and feedback: use the experience to retain, replace or redirect candidates.
3. Adjustment: change intensity/proportion and record each change and its basis.
4. Final candidates and record: organise reasons, feedback, confirmation and use context in a report.

A signal change does not establish liking. Subjective feedback can overturn a candidate.

When physiology and subjective experience differ, users need to be able to correct the choice instead of treating a candidate score as an answer they must accept.

Express preferences

The questionnaire collects use context, liking, intensity and sensitivity information as confirmable inputs.

Understand the process

The dashboard organises acquisition state, trends and experience stages, keeping indicators in context instead of displaying isolated numbers.

Inspect and adjust

Candidates include reasons. The console connects proportions, intensity, pause and adjustment history with feedback and the next operation.

Read the report

The report connects rationale, session feedback and final choices to support understanding and another choice.

1. 01Enter preferences
2. 02Inspect candidates and reasons
3. 03Adjust intensity or direction
4. 04Read report

The prototype remains in Chinese. Its charts and scores use simulated data.

Outputs include the cross-modal product proposal, experiment/evaluation design, seven-step service blueprint, five-layer system proposal, candidate/refinement mechanism and interaction/report prototypes. Basic-experience and deeper-customisation delivery options were also planned.

Next work needs consistent experimental conditions, actual acquisition/record alignment, device integration and task-based observation of whether users understand reasons, give feedback and make adjustments. Demonstration trends, scores and logs are not evidence of cross-modal recommendation performance.

This project connects technical possibilities to actions: what a device contributes, how an experiment produces evidence, how service supports the flow and how users keep control of the final choice.