This project deals with classification of an infant’s cry into hungry cry or spasmatic cry. Some infants have an early onset of idiopathic lung muscle spasm which results in improper supply of oxygen to the brain. When the cry of a baby is because of muscular spasms, there are some frequency ‘gaps’ in the spectrogram. The same is diagnosed with the help of an SVM and the cry is classified.
In this method, the Fourier transform is applied to a raw audio signal. The frequencies are binned between 20Hz to 20KHz (audible range). The sample is then fed to a pre-trained SVM for classification.
The jupyter-notebooks of the implementation and the sample recording can be found on my GitHub.