ASR Under Noise: Exploring Robustness for Sundanese and Javanese

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Pranida, Salsabila Zahirah, Airlangga, Muhammad Cendekia, Genadi, Rifo Ahmad, Shehata, Shady
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912617394601984
author Pranida, Salsabila Zahirah
Airlangga, Muhammad Cendekia
Genadi, Rifo Ahmad
Shehata, Shady
author_facet Pranida, Salsabila Zahirah
Airlangga, Muhammad Cendekia
Genadi, Rifo Ahmad
Shehata, Shady
contents We investigate the robustness of Whisper-based automatic speech recognition (ASR) models for two major Indonesian regional languages: Javanese and Sundanese. While recent work has demonstrated strong ASR performance under clean conditions, their effectiveness in noisy environments remains unclear. To address this, we experiment with multiple training strategies, including synthetic noise augmentation and SpecAugment, and evaluate performance across a range of signal-to-noise ratios (SNRs). Our results show that noise-aware training substantially improves robustness, particularly for larger Whisper models. A detailed error analysis further reveals language-specific challenges, highlighting avenues for future improvements
format Preprint
id arxiv_https___arxiv_org_abs_2509_25878
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ASR Under Noise: Exploring Robustness for Sundanese and Javanese
Pranida, Salsabila Zahirah
Airlangga, Muhammad Cendekia
Genadi, Rifo Ahmad
Shehata, Shady
Computation and Language
We investigate the robustness of Whisper-based automatic speech recognition (ASR) models for two major Indonesian regional languages: Javanese and Sundanese. While recent work has demonstrated strong ASR performance under clean conditions, their effectiveness in noisy environments remains unclear. To address this, we experiment with multiple training strategies, including synthetic noise augmentation and SpecAugment, and evaluate performance across a range of signal-to-noise ratios (SNRs). Our results show that noise-aware training substantially improves robustness, particularly for larger Whisper models. A detailed error analysis further reveals language-specific challenges, highlighting avenues for future improvements
title ASR Under Noise: Exploring Robustness for Sundanese and Javanese
topic Computation and Language
url https://arxiv.org/abs/2509.25878