On Out-of-Distribution Detection for Audio with Deep Nearest Neighbors

Zaharah Bukhsh; Aaqib Saeed

doi:10.1109/ICASSP49357.2023.10094846

On Out-of-Distribution Detection for Audio with Deep Nearest Neighbors

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

Out-of-distribution (OOD) detection is concerned with identifying data points that do not belong to the same distribution as the model's training data. For the safe deployment of predictive models in a real-world environment, it is critical to avoid making confident predictions on OOD inputs as it can lead to potentially dangerous consequences. However, OOD detection largely remains an under-explored area in the audio (and speech) domain. This is despite the fact that audio is a central modality for many tasks, such as speaker diarization, automatic speech recognition, and sound event detection. To address this, we propose to leverage feature-space of the model with deep k-nearest neighbors to detect OOD samples. We show that this simple and flexible method effectively detects OOD inputs across a broad category of audio (and speech) datasets. Specifically, it improves the false positive rate (FPR@TPR95) by 17% and the AUROC score by 7% than other prior techniques.

Original language	English
Title of host publication	ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Publisher	Institute of Electrical and Electronics Engineers
Pages	1-5
Number of pages	5
ISBN (Electronic)	978-1-7281-6327-7
DOIs	https://doi.org/10.1109/ICASSP49357.2023.10094846
Publication status	Published - 5 May 2023
Event	48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023 - Rhodes Island, Greece Duration: 4 Jun 2023 → 10 Jun 2023

Conference

Conference	48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023
Abbreviated title	ICASSP 2023
Country/Territory	Greece
City	Rhodes Island
Period	4/06/23 → 10/06/23

Keywords

Event detection
Neural networks
Training data
Predictive models
Signal processing
Feature extraction
Data models
out-of-distribution
nearest neighbors
uncertainty estimation
deep learning
speech
audio

Access to Document

10.1109/ICASSP49357.2023.10094846

Cite this

@inproceedings{cb496f725c8047f491c8f9db1fad04ca,

title = "On Out-of-Distribution Detection for Audio with Deep Nearest Neighbors",

abstract = "Out-of-distribution (OOD) detection is concerned with identifying data points that do not belong to the same distribution as the model's training data. For the safe deployment of predictive models in a real-world environment, it is critical to avoid making confident predictions on OOD inputs as it can lead to potentially dangerous consequences. However, OOD detection largely remains an under-explored area in the audio (and speech) domain. This is despite the fact that audio is a central modality for many tasks, such as speaker diarization, automatic speech recognition, and sound event detection. To address this, we propose to leverage feature-space of the model with deep k-nearest neighbors to detect OOD samples. We show that this simple and flexible method effectively detects OOD inputs across a broad category of audio (and speech) datasets. Specifically, it improves the false positive rate (FPR@TPR95) by 17\% and the AUROC score by 7\% than other prior techniques.",

keywords = "Event detection, Neural networks, Training data, Predictive models, Signal processing, Feature extraction, Data models, out-of-distribution, nearest neighbors, uncertainty estimation, deep learning, speech, audio",

author = "Zaharah Bukhsh and Aaqib Saeed",

year = "2023",

month = may,

day = "5",

doi = "10.1109/ICASSP49357.2023.10094846",

language = "English",

pages = "1--5",

booktitle = "ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

note = "48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023, ICASSP 2023 ; Conference date: 04-06-2023 Through 10-06-2023",

}

Bukhsh, Z & Saeed, A 2023, On Out-of-Distribution Detection for Audio with Deep Nearest Neighbors. in ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)., 10094846, Institute of Electrical and Electronics Engineers, pp. 1-5, 48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023, Rhodes Island, Greece, 4/06/23. https://doi.org/10.1109/ICASSP49357.2023.10094846

On Out-of-Distribution Detection for Audio with Deep Nearest Neighbors. / Bukhsh, Zaharah ; Saeed, Aaqib.
ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Institute of Electrical and Electronics Engineers, 2023. p. 1-5 10094846.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - On Out-of-Distribution Detection for Audio with Deep Nearest Neighbors

AU - Bukhsh, Zaharah

AU - Saeed, Aaqib

PY - 2023/5/5

Y1 - 2023/5/5

N2 - Out-of-distribution (OOD) detection is concerned with identifying data points that do not belong to the same distribution as the model's training data. For the safe deployment of predictive models in a real-world environment, it is critical to avoid making confident predictions on OOD inputs as it can lead to potentially dangerous consequences. However, OOD detection largely remains an under-explored area in the audio (and speech) domain. This is despite the fact that audio is a central modality for many tasks, such as speaker diarization, automatic speech recognition, and sound event detection. To address this, we propose to leverage feature-space of the model with deep k-nearest neighbors to detect OOD samples. We show that this simple and flexible method effectively detects OOD inputs across a broad category of audio (and speech) datasets. Specifically, it improves the false positive rate (FPR@TPR95) by 17% and the AUROC score by 7% than other prior techniques.

AB - Out-of-distribution (OOD) detection is concerned with identifying data points that do not belong to the same distribution as the model's training data. For the safe deployment of predictive models in a real-world environment, it is critical to avoid making confident predictions on OOD inputs as it can lead to potentially dangerous consequences. However, OOD detection largely remains an under-explored area in the audio (and speech) domain. This is despite the fact that audio is a central modality for many tasks, such as speaker diarization, automatic speech recognition, and sound event detection. To address this, we propose to leverage feature-space of the model with deep k-nearest neighbors to detect OOD samples. We show that this simple and flexible method effectively detects OOD inputs across a broad category of audio (and speech) datasets. Specifically, it improves the false positive rate (FPR@TPR95) by 17% and the AUROC score by 7% than other prior techniques.

KW - Event detection

KW - Neural networks

KW - Training data

KW - Predictive models

KW - Signal processing

KW - Feature extraction

KW - Data models

KW - out-of-distribution

KW - nearest neighbors

KW - uncertainty estimation

KW - deep learning

KW - speech

KW - audio

UR - https://www.scopus.com/pages/publications/85177564037

U2 - 10.1109/ICASSP49357.2023.10094846

DO - 10.1109/ICASSP49357.2023.10094846

M3 - Conference contribution

SP - 1

EP - 5

BT - ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

PB - Institute of Electrical and Electronics Engineers

T2 - 48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023

Y2 - 4 June 2023 through 10 June 2023

ER -

On Out-of-Distribution Detection for Audio with Deep Nearest Neighbors

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this