Speech Enhancement Noise Suppression Using DTLN

• Speech Enhancement: Tensorflow 2.x implementation of the stacked dual-signal transformation LSTM network (DTLN) for Noise Suppression.
• The code for training, inferring, and serving the DTLN model in Python is available in this repository. It also includes SavedModel, which contains pre-trained models that can be used as a starting point for your own projects.

Model Running Instructions

1. First you need to install python latest verion of python 3 and pip module.
2. After that you just need to go working Directory and run: pip run -r requirements.txt
3. After that you can run your JUPYTER NOTEBOOK and open file named: Model File.ipynb
4. All the model, training and testing instruction are given on that file.

Sample

I am showing you sample file befor noise reduction and after noise reduction.

• Normal Audio With Noises: Noisy Audio Input
• After Cleanig Noise: Clean Audio Output

Refrences

• Paper : Dual-Signal Transformation LSTM Network for Real-Time Noise Suppression
• Notes: Paper Notes