Gender Recognition of babies cry using Machine Learning Algorithms(Support Vector Machine and K-Nearest Neighbors) and Signal Processing(Fast Fourier Transform and Discrete Harley Transform)
Machine Learning Algorithms used for Prediction:
- Support Vector Machine
- K-Nearest Neighbor
Signal Pre-processing used:
- Fast Fourier Transform
- Discrete Hartley Transform (only uses the real values of Fast Fourier Transform)
Length of wavefile used in training and testing:
- 13000 data (you can use PCA for lessening the data)
- Use virtualenv in installing libraries and dependencies.
- Install dependencies and libraries via pip.
To install using pip
:
$ pip3 install requirements.txt
-datasets are on the Dataset folder. 16 sample cries for boy and girl are used (More dataset, more accurate)
- Training consists of boy and girl dataset.
- Either use Support Vector Machine or K-Nearest Neighbor for the Machine Learning Algorithm.
- Either use Fast Fourier Transform or Discrete Harley Transform for the Signal Processing.
Cut the wave files
$ python3 cutter.py
-cuts the wave files into desired length to be used as dataset or testing sample. Note: wave files must be in the same length.
Predict
$ python3 predict.py
-predict the testing sample wave file. Just input the location path of the wave file to be predicted
- Frequency Sampling
- Channels
- Complete Sampling Length
- Length in seconds of the wave file
- Time response of the program