[go: up one dir, main page]

0% found this document useful (0 votes)
157 views29 pages

CNN Lab Manual

This document discusses using deep learning for speech denoising. It details the steps to install Python, Jupyter Notebook, TensorFlow and other libraries. It then explains running a Jupyter notebook to train a 1D CNN and 2D CNN neural network models on noisy audio files. Screenshots show the training process and resulting waveforms are compared. SNR values of 17.64 and 14.63 are achieved for the 1D CNN and 2D CNN models respectively, indicating the noisy audio files were denoised. Block diagrams and algorithms for the 1D CNN and 2D CNN models are provided. Code is available at a GitHub link.

Uploaded by

Max Watson
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
157 views29 pages

CNN Lab Manual

This document discusses using deep learning for speech denoising. It details the steps to install Python, Jupyter Notebook, TensorFlow and other libraries. It then explains running a Jupyter notebook to train a 1D CNN and 2D CNN neural network models on noisy audio files. Screenshots show the training process and resulting waveforms are compared. SNR values of 17.64 and 14.63 are achieved for the 1D CNN and 2D CNN models respectively, indicating the noisy audio files were denoised. Block diagrams and algorithms for the 1D CNN and 2D CNN models are provided. Code is available at a GitHub link.

Uploaded by

Max Watson
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 29

DIGITAL SIGNAL PROCESSING

1. Title
2. Software/Platform used
3. Steps to install and run the code
4. Stepwise windows screenshots
5. Block Diagram
6. Algorithm used
7. Observation
8. Conclusion
9. Code Links
1) Title
Speech Denoising Using Deep learning

2) Software/Platform used
Coding: Python in Jupyter Notebook IDE
Neural Network made by: TensorFlow 2
Libraries: NumPy, Librosa, Soundfile etc.

3) Steps to install & run the code

1) Download python in given link: https://www.python.org/downloads/

• Install python
• Click all optional features including pip

• Python is installed
2) Now installing Jupyter notebook using pip in command prompt

python -m pip install jupyter

3) Install TensorFlow 2 using pip in command prompt:

pip install tensorflow

4) Installing other libraries like librosa, NumPy, soundfile etc using pip in command
prompt:

pip install librosa

pip install numpy

pip install soundfile

pip install matplotlib

pip install scipy

pip install wavefile

pip install ipython


Open File Containing 1D CNN jupyter with test & training files

In location bar, write jupyter notebook, then press “enter”. It opens terminal

This will open terminal and jupyter notebook


Open 1D-CNN.pynb file

Click in this block and press “shift + Enter” to run this block and continue till training of file
Run by pressing “shift + Enter “till this block starts training of audio files

Now, training starts


Here, training completes. So again, start running next blocks

Finally, we get SNR of 17.6417 by 1D CNN


After running jupyter notebook we get our output 1D CNN denoised audio wav file in our
given location
4) Stepwise windows screenshot
Working with jupyter notebook for speech denoising using deep learning.

Training files:
Training completes with 1000 epochs

SNR calculated after loading, training, processing

SNR: 17.641750
Noisy Input Files

Noisy audio input file 1

Noisy audio input file 2


Denoised audio output files by 1D CNN

Denoised audio output file 1 by 1D CNN

Denoised audio output file 2 by 1D CNN


Denoised audio output files by 1D CNN

Denoised audio file 1 by 2D CNN

Denoised audio file 2 by 2D CNN


Waveforms of audio files (1D CNN)

Waveform of input noisy audio file 1

Waveform of output denoised audio file 1 by 1D CNN


Waveforms of audio files (1D CNN)

Waveform of input noisy audio file 2

Waveform of output denoised audio file 2 by 1D CNN


Waveforms of audio files (2D CNN)

Waveform of input noisy audio file 1

2D CNN Denoised audio file 1


Waveforms of audio files (2D CNN)

Waveform of input noisy audio file 2

Waveform of denoised audio file 2 by 2D CNN


2D CNN training

Training started

Training ended
2D CNN SNR: 14.63251
Waveform of 2D CNN Denoised audio file 1

Waveform of 2D CNN Denoised audio file 2


5) Block Diagram

Figure 1: Block Diagram


6) Algorithm Used
Figure 2: 1D CNN by Blog/Good Audience
Figure 3: 2D CNN by Blog/Good Audience
9) Code Link: https://github.com/mohdshahbaz123/Speech-Denoising-using-Deep-
Learning

You might also like