| General Information |
|
| Repository Size and Activity |
|
| Contribution Statistics |
|
| Other Metrics |
|
| Application |
|
Note
Contributions to improve the completeness of this list are greatly appreciated. If you come across any overlooked papers, please feel free to create pull requests, open issues or contact me via email. Your participation is crucial to making this repository even better.
| Title | Repo | Paper |
|---|---|---|
| Diffusion-based Speech Enhancement in Matched and Mismatched Conditions using a Heun-based Sampler | ➖ |   |
| Unsupervised Speech Enhancement with Diffusion-based Generative Models |   |
  |
| Boosting Speech Enhancement with Clean Self-Supervised Features via Conditional Variational Autoencoders |  |
 |
| Diffusion-based Speech Enhancement with a Weighted Generative-Supervised Learning Loss |  |
  |
| AV2WAV: Diffusion-based Re-Synthesis from Continuous Self-Supervised Features for Audio-Visual Speech Enhancement | |
  |
| A Closer Look at Wav2vec2 Embeddings for On-Device Single-Channel Speech Enhancement | ➖ |   |
| GTCRN: A Speech Enhancement Model Requiring Ultralow Computational Resources |  |
 |
| Title | Repo | Paper |
|---|---|---|
StoRM: A Diffusion-based Stochastic Regeneration Model for Speech Enhancement and Dereverberation  |
 |
  |
| About | Repo |
|---|---|
| A Training Code Template for DNN-based Speech Enhancement |  |
| Title | Homepage | Paper |
|---|---|---|
| VoiceBank + DEMAND |  |
|
Code for SNR Calculation Examples
Note
It is important to understand that these ranges are approximate and can change depending on the specific application and quality requirements for speech enhancement. In addition, the perception of sound quality can be subjective and influenced by individual user preferences. In the context of the table provided, higher SNR values indicate better signal quality with less noise or interference.
| SNR Range | Description |
|---|---|
| Very Low SNR (<0 dB) | The noise power exceeds the signal power. The signal is almost entirely masked by noise, making recovery or processing extremely difficult. |
| Low SNR (0-10 dB) | Low SNR values indicate a significant presence of noise in the signal. Processing and perception of such a signal can be challenging but possible in some cases with specialized noise reduction algorithms. |
| Moderate SNR (10-20 dB) | These SNR values are considered acceptable for many signal processing applications. The signal contains a noticeable level of noise, but its quality is still sufficient for processing and perception. |
| High SNR (20-30 dB) | High SNR values indicate good signal quality with a low noise level. Such signals are easily processed and perceived by humans. |
| Very High SNR (>30 dB) | Very high SNR values correspond to signals with minimal noise levels. These signals are considered ideal for most signal processing applications. |
Code for SDR Calculation Examples
Note
It is important to understand that these ranges are approximate and can change depending on the specific application and quality requirements for speech enhancement. In addition, the perception of sound quality can be subjective and influenced by individual user preferences. In the context of the table provided, higher SNR values indicate better signal quality with less distortion.
| SDR Range | Description |
|---|---|
| Very Low SDR (<0 dB) | The distortion level exceeds the signal power. The signal is highly distorted, making it difficult to recover or process accurately. |
| Low SDR (0-5 dB) | Low SDR values indicate the presence of significant distortion relative to the signal power. Processing and perception of such a signal can be challenging, requiring specialized distortion reduction methods. |
| Moderate SDR (5-10 dB) | These SDR values indicate moderate distortion in the signal, which may degrade its quality, but still allows for some processing and perception. |
| High SDR (10-15 dB) | High SDR values indicate good signal quality relative to the level of distortion. Such signals are relatively clean and easy to process. |
| Very High SDR (>15 dB) | Very high SDR values correspond to signals with minimal distortion levels. These signals are considered ideal for most signal processing applications. |