https://github.com/machinelearningzuu/audio-recognition-with-cnn
#CNN #Signal Processing #Fourier Transform #TensorFlow
https://github.com/machinelearningzuu/audio-recognition-with-cnn
Last synced: 11 months ago
JSON representation
#CNN #Signal Processing #Fourier Transform #TensorFlow
- Host: GitHub
- URL: https://github.com/machinelearningzuu/audio-recognition-with-cnn
- Owner: machinelearningzuu
- Created: 2020-11-25T16:39:41.000Z (over 5 years ago)
- Default Branch: main
- Last Pushed: 2020-11-26T18:20:00.000Z (over 5 years ago)
- Last Synced: 2025-03-29T16:22:55.341Z (about 1 year ago)
- Language: Python
- Homepage:
- Size: 5.86 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Audio-Recognition-with-CNN
data set : http://storage.googleapis.com/download.tensorflow.org/data/mini_speech_commands.zip
### Data Preprocessing
#### 1) split data into train, test and val
rename the unzip folder (/mini_speech_commands) as /data/train. after executing **split_train_val_test** function it will create 2 more directories
(/data/val and /data/test) which contain train, test and validation splitted data from all classes.
#### 2) encode audio file to tensor
using **tf.io.read_file(filename)** each audio file needs to be convert into binary audio data file. then the binary file can be decode into a tensor using
**tf.audio.decode_wav(audio_binary)**
#### 3) Convert 1D signal into 2D
the decoded tensor is a vector. So this signal vector can be convered to signal matrix using fourier transform. Using Fourier transform it can convert time
series signal into frequency spectrum and which is 2 dimensional signal.