@inproceedings{mohammed-etal-2026-cadence,
title = "Cadence-Stratified Multimodal Perception for Real-Time Child{--}Robot Interaction",
author = "Mohammed, Marwan and
Zarrin, Javad and
ElSayed, Salma and
Kempe, Vera",
editor = "Ghaleb, Esam and
Takmaz, Ece and
Frassinelli, Diego and
Khaertdinov, Bulat and
Fern{\'a}ndez, Raquel and
Yumak, Zerrin and
Ozyurek, Asli",
booktitle = "Proceedings of the 1st Workshop on Multimodal Interaction in Face-to-Face Dialogue",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.mint-1.3/",
pages = "22--32",
ISBN = "979-8-89176-482-8",
abstract = "Child{--}robot interaction places real demands on perception systems: a robot engaged in play with a child cannot afford to miss the child{'}s focus of attention, yet running every perception module at every frame is computationally expensive on the kind of edge hardware that privacy-respecting robots must use. This paper addresses this tension by proposing a cadence-stratified multimodal perception pipeline, WordMate-CRI, in which each module runs at a rate that trades computation against staleness, informed by how long each cue remains stable for a potential interaction partner. Detection and tracking run continuously; action recognition uses rolling temporal buffers; emotion is refreshed periodically; and attributes such as age and gender are estimated once at the start of a session and cached. We evaluated WordMate-CRI through an LLM response-quality ablation study across six toy-play scenarios and a latency benchmark on Jetson AGX Orin hardware. The ablation shows that VLM enrichment and object evidence are the dominant grounding signals, with WordMate-CRI achieving a mean scene-grounded score of 0.967. On live Jetson input, cadence-based scheduling raises throughput from 4.3 to 18.2 frames per second (FPS), clearing the 15 FPS deployment target, while reducing WordMate-CRI latency from 6411.5 to 5844.5 ms. Taken together, the results make a practical case that deciding when each perception module should run matters just as much as deciding what it should recognise. This optimisation is important for systems aiming to produce contingent and timely child-directed speech in multi-modal contexts."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mohammed-etal-2026-cadence">
<titleInfo>
<title>Cadence-Stratified Multimodal Perception for Real-Time Child–Robot Interaction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Marwan</namePart>
<namePart type="family">Mohammed</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Javad</namePart>
<namePart type="family">Zarrin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Salma</namePart>
<namePart type="family">ElSayed</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vera</namePart>
<namePart type="family">Kempe</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 1st Workshop on Multimodal Interaction in Face-to-Face Dialogue</title>
</titleInfo>
<name type="personal">
<namePart type="given">Esam</namePart>
<namePart type="family">Ghaleb</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ece</namePart>
<namePart type="family">Takmaz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Diego</namePart>
<namePart type="family">Frassinelli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bulat</namePart>
<namePart type="family">Khaertdinov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Raquel</namePart>
<namePart type="family">Fernández</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zerrin</namePart>
<namePart type="family">Yumak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Asli</namePart>
<namePart type="family">Ozyurek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-482-8</identifier>
</relatedItem>
<abstract>Child–robot interaction places real demands on perception systems: a robot engaged in play with a child cannot afford to miss the child’s focus of attention, yet running every perception module at every frame is computationally expensive on the kind of edge hardware that privacy-respecting robots must use. This paper addresses this tension by proposing a cadence-stratified multimodal perception pipeline, WordMate-CRI, in which each module runs at a rate that trades computation against staleness, informed by how long each cue remains stable for a potential interaction partner. Detection and tracking run continuously; action recognition uses rolling temporal buffers; emotion is refreshed periodically; and attributes such as age and gender are estimated once at the start of a session and cached. We evaluated WordMate-CRI through an LLM response-quality ablation study across six toy-play scenarios and a latency benchmark on Jetson AGX Orin hardware. The ablation shows that VLM enrichment and object evidence are the dominant grounding signals, with WordMate-CRI achieving a mean scene-grounded score of 0.967. On live Jetson input, cadence-based scheduling raises throughput from 4.3 to 18.2 frames per second (FPS), clearing the 15 FPS deployment target, while reducing WordMate-CRI latency from 6411.5 to 5844.5 ms. Taken together, the results make a practical case that deciding when each perception module should run matters just as much as deciding what it should recognise. This optimisation is important for systems aiming to produce contingent and timely child-directed speech in multi-modal contexts.</abstract>
<identifier type="citekey">mohammed-etal-2026-cadence</identifier>
<location>
<url>https://aclanthology.org/2026.mint-1.3/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>22</start>
<end>32</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Cadence-Stratified Multimodal Perception for Real-Time Child–Robot Interaction
%A Mohammed, Marwan
%A Zarrin, Javad
%A ElSayed, Salma
%A Kempe, Vera
%Y Ghaleb, Esam
%Y Takmaz, Ece
%Y Frassinelli, Diego
%Y Khaertdinov, Bulat
%Y Fernández, Raquel
%Y Yumak, Zerrin
%Y Ozyurek, Asli
%S Proceedings of the 1st Workshop on Multimodal Interaction in Face-to-Face Dialogue
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-482-8
%F mohammed-etal-2026-cadence
%X Child–robot interaction places real demands on perception systems: a robot engaged in play with a child cannot afford to miss the child’s focus of attention, yet running every perception module at every frame is computationally expensive on the kind of edge hardware that privacy-respecting robots must use. This paper addresses this tension by proposing a cadence-stratified multimodal perception pipeline, WordMate-CRI, in which each module runs at a rate that trades computation against staleness, informed by how long each cue remains stable for a potential interaction partner. Detection and tracking run continuously; action recognition uses rolling temporal buffers; emotion is refreshed periodically; and attributes such as age and gender are estimated once at the start of a session and cached. We evaluated WordMate-CRI through an LLM response-quality ablation study across six toy-play scenarios and a latency benchmark on Jetson AGX Orin hardware. The ablation shows that VLM enrichment and object evidence are the dominant grounding signals, with WordMate-CRI achieving a mean scene-grounded score of 0.967. On live Jetson input, cadence-based scheduling raises throughput from 4.3 to 18.2 frames per second (FPS), clearing the 15 FPS deployment target, while reducing WordMate-CRI latency from 6411.5 to 5844.5 ms. Taken together, the results make a practical case that deciding when each perception module should run matters just as much as deciding what it should recognise. This optimisation is important for systems aiming to produce contingent and timely child-directed speech in multi-modal contexts.
%U https://aclanthology.org/2026.mint-1.3/
%P 22-32
Markdown (Informal)
[Cadence-Stratified Multimodal Perception for Real-Time Child–Robot Interaction](https://aclanthology.org/2026.mint-1.3/) (Mohammed et al., MINT 2026)
ACL