Trigger Sound Suppression for Misophonia

Vaishnavi Vidyasagar; Jasmine Zhang; Mahima Uliyar; Seunghyun Oh; Emily Catherine Gates; Mark Zachary Rosenthal; Shyamnath Gollakota

cs.SDSep 28, 2026

Trigger Sound Suppression for Misophonia

Authors: Vaishnavi Vidyasagar, Jasmine Zhang, Mahima Uliyar, Seunghyun Oh, Emily Catherine Gates, Mark Zachary Rosenthal, Shyamnath Gollakota

Organizations: University of Washington · Duke Center for Misophonia and Emotion Regulation

Abstract

Misophonia, a disorder of decreased tolerance to specific sounds, affects 5-20% of the population, yet sufferers have no good options: therapy helps a minority, and earplugs or noise cancellation silence everything. We present a study for neural trigger sound suppression for misophonia, selectively removing trigger sounds. We curate a dataset covering the 10 most common trigger classes. Using streaming dual-path networks operating on 6 ms audio chunks, we explore both one-hot and multi-hot-conditioned models that suppress 1-3 triggers from the acoustic scene. We validate our model outputs in a listening study with 30 adults with clinically elevated misophonia impairment. Participants reported significantly lower distress and arousal, and improved valence, for suppressed audio.

Figures & tables

Trigger	Chewing	Cough	Drinking	Gum pop	Heavy breath	Lip smack	Sneezing	Sniffling	Tapping	Throat clearing	Total
FSD50K	231	-	-	-	-	-	-	-	-	-	231
CoughVID	-	1,187	-	-	-	-	-	-	-	-	1,187
ESC-50	-	40	40	-	-	-	20	-	-	-	100
MATA	-	-	34	-	-	21	12	28	-	-	95
VocalSound	-	-	-	-	-	-	332	-	-	3,504	3,836
Freesound	-	-	-	4	-	-	38	8	149	-	199

Table 1 : Number of audio clips per trigger class in our dataset.

Trigger Class	SI-SNRi (dB)	SNRi (dB)
Chewing	$11.88_{\pm 4.36}$	$13.18_{\pm 4.26}$
Cough	$15.79_{\pm 5.47}$	$15.91_{\pm 5.30}$
Drinking	$10.73_{\pm 3.52}$	$11.74_{\pm 3.48}$
Gum popping	$13.29_{\pm 4.07}$	$13.85_{\pm 3.96}$
Heavy Breathing	$12.48_{\pm 3.79}$	$13.31_{\pm 3.81}$
Lip smacking	$14.35_{\pm 5.27}$	$15.07_{\pm 5.05}$

Table 2 : Performance of a single-trigger multi-class model trained across all 10 trigger classes, evaluated with single-trigger mixtures and a single output channel.

		# selected trigger classes
Model	Metric	1	2	3
Ours	SI-SNRi	$19.86_{\pm 5.82}$	$18.17_{\pm 5.23}$	$16.72_{\pm 5.10}$
	SNRi	$19.99_{\pm 5.47}$	$18.69_{\pm 4.82}$	$17.78_{\pm 4.43}$
Waveformer	SI-SNRi	$12.92_{\pm 5.06}$	$12.30_{\pm 4.17}$	$11.90_{\pm 3.88}$
	SNRi	$13.67_{\pm 4.87}$	$13.76_{\pm 4.07}$	$14.07_{\pm 3.93}$

Table 3: Multi-trigger suppression model with 0.496M parameters, compared to 2.020M for Waveformer baseline.

( $L_{C}/L_{B}/L_{F}$ , ms)	Latency	SNRi (dB)	SI-SNRi (dB)
6/6/4	10 ms	$19.99_{\pm 5.47}$	$19.86_{\pm 5.82}$
8/4/4	12 ms	$17.08_{\pm 5.02}$	$16.76_{\pm 5.25}$
4/12/0	4 ms	$13.38_{\pm 4.84}$	$12.55_{\pm 5.08}$

Table 4 : Ablation study over STFT parameters for the multi-trigger model, evaluated with a single trigger queried.

Figure 1 : Time-domain waveforms for gum popping suppression: input mixture, ground truth, and model output.

Measure	Original	Output	$\Delta$ [95% CI]
Distress (1–10)	$7.68\pm 1.60$	$5.31\pm 2.48$	$-2.37$ [1.45, 3.28]
Arousal (1–9)	$7.02\pm 1.42$	$4.89\pm 2.19$	$-2.13$ [1.36, 2.91]
Valence (1–9)	$4.42\pm 2.75$	$6.01\pm 1.65$	$+1.59$ [1.05, 2.13]

Table 5 : Affective response to raw audio vs. model output (30 participants). Lower is better for distress and arousal; higher better for valence. All paired

t

-tests

p<.001

Explore similar work

A Neuromorphic Trigger for Efficient Audio Event Detection
Jun 16, 2026Benjamin Hatton, Oliver Rhodes, Luca PeresNeuromorphic Computing Spiking Neural Networks
ICNS, University of Manchester Manchester, UK
Causal Analysis and Mitigation of Spurious Onsets in Full-Duplex Speech LLMs
Sep 11, 2026Kento NishiSpeech Language Models Full-Duplex Speech Models
Massachusetts Institute of Technology · Comcast Applied AI Research, Speech AI Team
Focus Then Listen: An Empirical Study of Plug-and-Play Audio Enhancer for Noise-Robust Large Audio Language Models
Mar 5, 2026Han Yin, Yang Xiao, Younghoo Kwon +2Large Audio Language Models Audio Understanding
School of Electrical Engineering, KAIST, Daejeon, Republic of Korea · University of Melbourne, Australia

Jun 16, 2026, cs.SDJ/K move · Enter open · S save

A Neuromorphic Trigger for Efficient Audio Event Detection

Benjamin Hatton, Oliver Rhodes, Luca Peres

ICNS, University of Manchester Manchester, UK

Neuromorphic Computing Spiking Neural Networks Audio Understanding

Open paper AskPDF arXiv

Efficient processing of continuous audio streams remains a key challenge for real-time and resource-constrained systems. This paper introduces a neuromorphic trigger for audio event detection, based on a spiking neural network (SNN) that selectively gates input to downstream models. The proposed neuromorphic trigger acts as a flexible low-cost front-end, identifying salient audio segments and enabling these to be processed by a more computationally intensive model for tasks such as classification. The trigger is implemented as a lightweight fully connected SNN using a close-open filter for postprocessing, and is evaluated on two representative tasks: Anomalous Sound Detection (ASD) and Sound Event Detection (SED). For ASD, the trigger achieves a one-second segment-based F1 score of 0.97 on a class-agnostic form of the URBAN-SED dataset, demonstrating high reliability in identifying relevant audio regions. For SED, the trigger is combined with the Dang classifier on the DCASE 2017 Challenge Task 2 dataset, showing a potential

42.6\times

reduction in FLOPs while reducing the lower bound of the event-based error rate from 0.41 to 0.25. These results highlight the potential of neuromorphic triggers as real-time, energy-efficient front-end filters, enabling substantial reductions in computational cost.

Abstract

Explore similar work

A Neuromorphic Trigger for Efficient Audio Event Detection

Causal Analysis and Mitigation of Spurious Onsets in Full-Duplex Speech LLMs

Focus Then Listen: An Empirical Study of Plug-and-Play Audio Enhancer for Noise-Robust Large Audio Language Models