mlsanityData to 19 Oct 2026

DualSpecSE: A Dual-Path Speech Enhancement Network Integrating Mel and Complex Spectrograms

arXiv 2609.13911 ↗First posted 12 Sep 2026eess.AS cs.SDcited by 0

On the map: Speech and audio language models → Speech enhancement and noise suppression

Abstract (arXiv)

In this paper, we propose DualSpecSE, a speech enhancement framework that jointly models Mel-spectrogram and complex spectrogram in a dual-path architecture for improved ASR performance and higher-quality speech reconstruction. The Mel branch learns coarse-grained acoustic representations and produces enhanced Mel-spectrograms for direct ASR usage, while the complex branch refines fine-grained spectral details for high-fidelity waveform reconstruction. Built upon the cross-band and narrow-band blocks from CleanMel, DualSpecSE introduces an interaction module and a fusion module to enable effective information exchange between the two branches. The model simultaneously outputs enhanced Mel and complex spectrogram without requiring a pretrained vocoder. Experimental results demonstrate consistent improvements in speech fidelity, perceptual quality, and ASR performance. Codes and audio samples are available.

Key facts. No citing papers in the map yet (data through 19 Oct 2026). Placed in Speech enhancement and noise suppression, one of 11 directions in Speech and audio language models. Builds on 22 of its 31 references that are in the map. First posted on arXiv 12 Sep 2026.

Builds on

22 of 31 references.

Closest ancestors · 8

Same direction, newest first

  1. 2025A Composite Predictive-Generative Approach to Monaural Universal Speech Enhancement30 May
  2. 2025Mel-McNet: A Mel-Scale Framework for Online Multichannel Speech Enhancement26 May
  3. 2025CleanMel: Mel-Spectrogram Enhancement for Improving Both Speech Quality and ASR27 Feb
  4. 2024EARS: An Anechoic Fullband Speech Dataset Benchmarked for Speech Enhancement and Dereverberation10 Jun
  5. 2023MP-SENet: A Speech Enhancement Model with Parallel Denoising of Magnitude and Phase Spectra23 May
  6. 2023DeepFilterNet: Perceptually Motivated Real-Time Speech Enhancement14 May
  7. 2022StoRM: A Diffusion-based Stochastic Regeneration Model for Speech Enhancement and Dereverberation22 Dec
  8. 2022CMGAN: Conformer-Based Metric-GAN for Monaural Speech Enhancement22 Sep

Other references · 14

  1. 2023ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoderscited by 737
  2. 2020DNSMOS: A Non-Intrusive Perceptual Objective Speech Quality metric to evaluate Noise Suppressorscited by 356
  3. 2020DCCRN: Deep Complex Convolution Recurrent Network for Phase-Aware Speech Enhancementcited by 284
  4. 2024Emilia: An Extensive, Multilingual, and Diverse Speech Dataset for Large-Scale Speech Generationcited by 268
  5. 2017SEGAN: Speech Enhancement Generative Adversarial Networkcited by 262
  6. 2020The INTERSPEECH 2020 Deep Noise Suppression Challenge: Datasets, Subjective Testing Framework, and Challenge Resultscited by 230
  7. 2020FullSubNet: A Full-Band and Sub-Band Fusion Model for Real-Time Single-Channel Speech Enhancementcited by 121
  8. 2022TF-GridNet: Making Time-Frequency Domain Models Great Again for Monaural Speaker Separationcited by 76
  9. 2017A Hybrid DSP/Deep Learning Approach to Real-Time Full-Band Speech Enhancementcited by 67
  10. 2020A Perceptually-Motivated Approach for Low-Complexity, Real-Time Enhancement of Fullband Speechcited by 53
  11. 2023SpatialNet: Extensively Learning Spatial Information for Multichannel Joint Speech Separation, Denoising and Dereverberationcited by 39
  12. 2022Speech Denoising in the Waveform Domain with Self-Attentioncited by 34
2 more
  1. 2023FSD: An Initial Chinese Dataset for Fake Song Detectioncited by 13
  2. 2024FreeV: Free Lunch For Vocoders Through Pseudo Inversed Mel Filtercited by 8

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Mel-FullSubNet: Mel-Spectrogram Enhancement for Improving Both Speech Quality and ASR
  2. 2023speech and noise dual-stream spectrogram refine network with speech distortion loss for robust speech recognition
  3. 2026Dual-View Predictive Diffusion: Lightweight Speech Enhancement via Spectrogram-Image Synergy
  4. 2024A Mel Spectrogram Enhancement Paradigm Based on CWT in Speech Synthesis
  5. 2024A Dual-Branch Parallel Network for Speech Enhancement and Restoration
  6. 2023D2Former: A Fully Complex Dual-Path Dual-Decoder Conformer Network using Joint Complex Masking and Complex Spectral Mapping for Monaural Speech Enhancement
  7. 2026U-PAST: A Phase-Aware Audio Spectrogram Transformer-U-Net for Single-Channel Speech Enhancement
  8. 2025Lightweight Front-end Enhancement for Robust ASR via Frame Resampling and Sub-Band Pruning