The argument
behind the object.

Tactons / Brewster & Brown, 2004

Structured tactile messages can use rhythm, duration and other parameters. For Fermata, this motivates a small vocabulary whose distinguishability must be tested. It does not establish that this particular wearable’s cues are understandable.

Musical Haptic Wearables / Turchet & Barthet, 2019

A co-design approach with electronic musicians connects haptic wearables to creative communication. Fermata carries forward the need to design for performers’ attention and shared activity. Placement, training and performance context belong in the evaluation.

MiMU / connect the action to its mapping

The official documentation distinguishes movement and events, and makes input/output mapping explicit. The transferable lesson is to show how a bodily action becomes an intentional signal. Fermata uses cues between wearers; it is not presented as a MIDI glove or music generator.

Shape Haptics / let the mechanism carry the image

Clement Zheng’s project page ties mechanism, applications, research and a usable design tool together. For this case study, large geometry-derived hardware views and local explanations carry the technical story.

Adafruit CLUE BLE MIDI Glove / connect code and use

The guide brings the parts, code, display and control mappings into one sequence. Fermata follows that documentation principle with a component-role table, real code excerpts and the actual recording UI.

Jazz Hand / document a specific failure

Sean Johnson describes signal smoothing, debouncing and wire routing in a wearable music interface. The lesson for the portfolio is editorial: describe the observed problem, the change and the remaining limit, with source evidence beside them.

The Malbody Centre / keep the body in the test

Marije Baalman’s prototype notes show that garments and skin contact can alter haptic perception. Fermata’s proposed evaluation therefore includes contact, fit, ordinary movement and interpretation.

Slap Snap Tap / explain through actions

MIT Media Lab’s overview describes sensor straps, physical programming and bodily action. This supports a clear action sequence as a documentation technique; no educational outcome is borrowed for Fermata.

NYU ITP / separate the case study from the process log

The documentation guide asks for the participant’s experience, a system diagram, circuit documentation, readable source and the difficulties encountered. This website is an edited path through the archive, while the local research notes retain the detailed source inventory.

Image provenance

The hero is a new imagegen visualization guided by the author’s actual prototype photograph. The on-wrist and open-case photos are documentary. The Blender images distinguish imported carrier/CAD geometry from newly reconstructed functional modules, with illustrative materials and no baked portfolio text. The exploded view combines the v1.3.2 base and v1.3.1 retained lid, with user-reported battery and LRA dimensions. The light strip sits on the original archived bridge, whose two feet meet the shell ledges. A separate cutaway shows the diffuser and lid window. The v4 bridge restores the nominal mounting height from the archive; LED thickness, adhesive and the optical gap remain reconstruction assumptions. Driver size, LED spacing and lead paths remain approximate; the model does not establish physical fit.

Recorded signal

Session 20261019_142008_test_cc691f, A v326 / recognizer v322, USB, desktop Test. One successful group was selected for the diagram: 729 samples over seven seconds, 53 masked samples. It illustrates distinct events without representing overall performance. Device ACK is logged on A; markers include human response delay.

Training, runtime and recording

Raw records place acceleration before angular velocity; the model expects gyro first, in physical units. The implementation normalizes time to 128 positions and uses a fixed gyro scale of 400. Session splits, group weighting and a fixed training-only hard-negative pass reduce misleading clip-level comparisons. Current A/B firmware uses v322; v325 remains a candidate.

The server stores device microseconds, host monotonic receipt time and browser marker time separately. A gap above 30 ms breaks a model window; a reset or interruption above one second saves and ends the recording. Role selection and panel feedback/calibration commands are locked during capture; physical buttons remain active.

A supported light assembly

The v4 assembly and local section preserve the archived bridge, diffuser, shell and lid geometry. Eighteen sampled bridge-foot contact positions agree with the original shell ledge height. This verifies the nominal CAD relationship, not printed fit or light uniformity. Driver size, wiring, LED package thickness and adhesive remain approximate.

Three views of the workbench

Observe and Mark show enlarged viewports into the unchanged 906 × 817 pixel UI capture. Trace is a rendered event ledger from a separate retained A USB session. The interactive waveform uses all 729 raw samples in the selected excerpt and a fixed ±1500 °/s vertical range. Two human cycle-end markers are present; an explicit group-onset marker is absent. A's motor and matching ACK are logged, but B's playback and a human response are not in this record.

Three bounded engineering cases

The window case compares historical replay after both labels and model fitting changed. The motor-mask case retains a null result: masked and unmasked replay logs were identical in the historical v316 DTW-era record. The delivery case separates v309 ACK_NO_HAPTIC evidence from seven later v326 physical button checks. Human confirmation describes the set of button tests, not individually timestamped sensations.

Demonstration / filming preparation

The four-step region is a planned filming storyboard, not recorded footage or an observed successful conversation. The requested film has not yet been shot. The local production guide specifies one continuous master take, wrist and reaction closeups, cue confirmation, firmware and take logs, and an edit sequence. Once captured, the central stage can become a real video while retaining the original take and session provenance.