Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 1509.05254

Post-processing speech recordings during MRI

Juha Kuortti·Jarmo Malinen·Antti Ojalammi
View original source ↗Read source document ↗Export metadata ↓

Abstract

We discuss post-processing of speech that has been recorded during Magnetic Resonance Imaging (MRI) of the vocal tract. Such speech recordings are contaminated by high levels of acoustic noise from the MRI scanner. Also, the frequency response of the sound signal path is not flat as a result of severe restrictions on recording instrumentation due to MRI technology. The post-processing algorithm for noise reduction is based on adaptive spectral filtering. The speech material consists of samples of prolonged vowel productions that are used for validation of the post-processing algorithm. The comparison data is recorded in anechoic chamber from the same test subject. Formant analysis is carried out for the post-processed speech and the comparison data. Artificially noise-contaminated vowel samples are used for validation experiments to determine performance of the algorithm where using true data would be difficult. The properties of recording instrumentation or the post-processing algorithm do not explain the consistent frequency dependent discrepancy between formant data from experiments during MRI and in anechoic chamber. It is shown that the discrepancy is statistically significant, in particular, where it is largest at 1 kHz and 2 kHz. The reflecting surfaces of the MRI head and neck coil are suspected to change the speech acoustics which results in "external formants" at these frequencies. However, the role of test subject adaptation to noise and constrained space acoustics during an MRI examination cannot be ruled out.

Explore related subjects

cs.SDphysics.bio-ph

Keep this discovery

BibTeXRIS

Juha Kuortti, Jarmo Malinen, Antti Ojalammi. 2015-09-17. Post-processing speech recordings during MRI. https://arxiv.org/abs/1509.05254

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Xiaomi-CocktailASR-1 Technical Report

Recently, large language model (LLM) based ASR models have achieved significant progress, yet they generally lack support for multi-speaker scenarios, where the cocktail party problem remains a critical bottleneck for further advancing ASR. Existing TS-ASR methods, including end-to-end architectures with speaker embeddings and latest LLM-based explorations suffer from degraded single-speaker performance and the inability to reject when the target speaker is absent. In this paper, we propose Xiaomi-CocktailASR-1, an LLM-based end-to-end TS-ASR architecture. By utilizing reference speech as voiceprint prompts, it directly transcribes the target speaker's speech without requiring speech separation. Xiaomi-CocktailASR-1 maintains competitive performance in single-speaker scenarios, comparable to mainstream ASR models. It also features a negative sample rejection capability, outputting empty text when the target speaker is absent from the mixed speech. Additionally, Xiaomi-CocktailASR-1 supports a Chain-of-Thought (CoT) reasoning mode to provide explicit reasoning steps. Extensive experiments on various synthetic and real-world multispeaker benchmarks demonstrate that Xiaomi-CocktailASR-1 achieves state-of-the-art performance, effectively addressing the cocktail party problem through a unified architecture that balances multispeaker and single-speaker recognition accuracy, along with rejection capability.

cs.SD↗
arXiv2026

EConv-TasNet: Efficient Conv-TasNet for Effective Speech Separation

Conv-TasNet has served as a strong baseline for time-domain speech separation, and many studies have extended it with advanced architectures such as dual-path networks, U-Nets, and attention mechanisms. However, these methods often introduce high computational cost and complexity, limiting their deployment in resource-constrained scenarios. To address this issue, we propose eConv-TasNet, an efficient variant of Conv-TasNet that improves both effectiveness and efficiency without relying on resource-intensive modules. The proposed model consists of a group-wise early-splitting (GES) module and a multi-group feature aggregation (MGFA) module. GES generates discriminative speaker embeddings at intermediate stages, while MGFA progressively aggregates these group-level representations for refined mask estimation. Experimental results show that eConv-TasNet reduces model size by 22.4%, accelerates inference by 18.9%, and improves SI-SNRi by 14.0%-28.0% across three public benchmarks. Moreover, it achieves competitive performance compared with state-of-the-art methods while requiring significantly fewer parameters and lower inference cost. These results demonstrate a favorable efficiency-effectiveness trade-off for edge deployment.

cs.SD↗