# Spectral feature mapping with mimic loss for robust speech recognition

arXiv: https://arxiv.org/abs/1803.09816 · HTML: https://mlsanity.com/paper/1803.09816/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 26 Mar 2018.

## Built on by (most cited first, 8 of 8)

- [SpeechBrain: A General-Purpose Speech Toolkit](https://mlsanity.com/paper/2106.04624/) (2021)
- [Conditional Diffusion Probabilistic Model for Speech Enhancement](https://mlsanity.com/paper/2202.05256/) (2022)
- [MetricGAN+: An Improved Version of MetricGAN for Speech Enhancement](https://mlsanity.com/paper/2104.03538/) (2021)
- [Discretization and Re-synthesis: an alternative method to solve the Cocktail Party Problem](https://mlsanity.com/paper/2112.09382/) (2021)
- [MetricGAN+/-: Increasing Robustness of Noise Reduction on Unseen Data](https://mlsanity.com/paper/2203.12369/) (2022)
- [Linguistic Knowledge Transfer Learning for Speech Enhancement](https://mlsanity.com/paper/2503.07078/) (2025)
- [Transcription-Free Fine-Tuning of Speech Separation Models for Noisy and Reverberant Multi-Speaker Automatic Speech Recognition](https://mlsanity.com/paper/2406.08914/) (2024)
- [Joint Learning using Mixture-of-Expert-Based Representation for Speech Enhancement and Robust Emotion Recognition](https://mlsanity.com/paper/2509.08470/) (2025)

## Abstract (arXiv)

For the task of speech enhancement, local learning objectives are agnostic to phonetic structures helpful for speech recognition. We propose to add a global criterion to ensure de-noised speech is useful for downstream tasks like ASR. We first train a spectral classifier on clean speech to predict senone labels. Then, the spectral classifier is joined with our speech enhancer as a noisy speech recognizer. This model is taught to imitate the output of the spectral classifier alone on clean speech. This textitmimic loss is combined with the traditional local criterion to train the speech enhancer to produce de-noised speech. Feeding the de-noised speech to an off-the-shelf Kaldi training recipe for the CHiME-2 corpus shows significant improvements in WER.