← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

MorsEar: Toward Generalizable Low-Resource Covert Messaging via Earable based Inertial Sensing

Garvit Chugh , Indrajeet Ghosh , Nirmalya Roy , Sandip Chakraborty , Suchetana Chakraborty

DOI PDF Program page Page JSON

MorsEar is a credible CHI systems contribution: it turns an earable IMU into a covert text-entry channel with a coherent interaction grammar and an on-device decoder, and it backs that claim with multi-context user data. The main caveat is that the system’s promise depends on Morse learning and a bounded evaluation scope.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

MorsEar’s strongest contribution is architectural rather than merely incremental. The paper does not just bolt Morse onto a wearable; it argues that commodity earable IMUs can support covert, eyes-free, voice-free language-level interaction if the sensing, segmentation, and decoding stack are designed together. The novelty claim is grounded in the combination of physics-aware preprocessing, tempo-adaptive rolling buffers, and an on-device decoder with autocorrect, which is a coherent answer to the practical problem of turning noisy near-ear micro-gestures into usable text entry. That is a meaningful departure from the common-sense expectation that text entry in constrained environments must rely on touch, speech, or specialized hardware. The validation is also reasonably broad for CHI: the paper reports a 24-participant study across Silent, Cafe, and Metro, includes LOSO-style generalization, latency measurements, and a short deployment/learning component. Those data support the claim that the system is usable and responsive in multiple contexts, and that autocorrect materially improves performance. At the same time, the paper is careful enough to acknowledge important limits: Morse has a learning barrier, the deployment evidence is short-term, and the ecological coverage is not exhaustive. So the right reading is not that MorsEar solves covert text entry universally, but that it establishes a credible design family for low-resource earable messaging and demonstrates that the approach can work beyond toy command sets. That makes it a solid honorable-mention-level systems paper with clear technical ambition and a sensible, if bounded, empirical basis.

What Changed

Canon before

Prior wearable covert text-entry systems typically depended on custom sensors, constrained vocabularies, or discrete command sets rather than a generalizable language-level earable IMU-only Morse interaction.

Departure from common sense

The paper’s core move is to treat commodity earable IMUs as a viable substrate for covert, eyes-free, voice-free text entry by encoding near-ear micro-gestures into Morse and then into unrestricted character composition. That is a non-obvious departure from the usual assumption that practical text entry needs touch, speech, or specialized hardware.

Actual novelty

The novelty is not just “Morse on a wearable,” but an ear-native interaction stack: physics-aware preprocessing, a tempo-adaptive grammar with rolling symbol/character/word buffers, and an on-device decoder with lightweight autocorrect. Together these are presented as the mechanism that makes IMU-only covert messaging scale from discrete commands to language-level interaction.

Evidence

The paper combines a system contribution with empirical validation. It reports a 24-participant study across Silent, Cafe, and Metro, includes LOSO generalization, latency measurements, and a 7-day learning/usability component. The evidence supports the claim that the system works under multiple contexts and that autocorrect improves performance, while also showing the learning burden of Morse.

Limits

Method limits

The evaluation is strong for a CHI systems paper but still bounded: the study size is modest, the interaction vocabulary is Morse-based, and the reported learning results reflect short-term familiarization rather than long-term mastery. The system’s performance is tied to the tested gesture set, decoder design, and the specific contexts sampled in the study.

Deployment limits

Deployment is limited by the need for user familiarization with Morse timing and rhythm, by the reliance on near-ear micro-gestures that may not be equally practical in all settings, and by the fact that the reported deployment evidence covers early viability rather than sustained real-world use over long periods.

Boundary conditions

The approach is most plausible for low-resource, covert, eyes-free, voice-free messaging scenarios where brief practice is acceptable and where users can perform near-ear gestures reliably. It is less directly supported for long-term fluent use, unconstrained daily activity, or settings where gesture execution or timing consistency is degraded.

Manual check: false

Position in field

This sits at the intersection of wearable sensing, accessible text entry, and covert communication. Its contribution is a generalizable interaction architecture for earable IMU-based messaging rather than a narrow command recognizer, and its CHI value is in showing that a compact sensing modality can support language-level interaction with real-time feedback.

Abstract

Silent, eyes-free text entry remains challenging when speech and touch are impractical. Prior wearable systems required custom sensors or limited users to a small vocabulary. We present MorsEar, an IMU-only earable framework that maps near-ear micro-gestures such as taps for dot/dash; slide/pull/circle for space/delete/send, into character-level Morse, enabling unrestricted composition with a compact lexicon for lightweight on-device autocorrect. The result is a low-bandwidth, reduced-exposure communication channel that works eyes-free and voice-free in accessibility scenarios, silent zones, and constrained environments. MorsEar infers words using a physics-aware preprocessing stack and a compact CNN, feeding a tempo-adaptive segmentation with rolling buffers; an on-device decoder provides real-time feedback entirely on-phone. In a 24-participant study (with four accessibility users) across Silent, Cafe, and Metro, MorsEar achieved CER 7.3% and WER 12.5% → 7.8% (Autocorrect), with median WPM of 9.3/9.1/5.8, respectively. Similar to other accessibility-oriented encodings such as Braille, Morse requires a brief familiarization period to learn the timing and rhythm of dots and dashes; after which, MorsEar shows that commodity earable IMUs can support discreet, low-exposure text entry that scales beyond discrete commands to language-level interaction.