An approach to finding the most suitable algorithm for noise reduction in a Neonatal cry signal

In this paper noise reduction in audio recording is a challenging work, especially when environment is having many other machineries. In this research hospital environment is a subject matter. The pre-processing of the cry signals with the background noise of Neonatal Intensive Care Units (NICUs) is challenging. These noises affect the accuracy of machine-learning techniques because cry recordings of healthy neonates are not from the same environment. This paper compares different techniques of noise reduction. The most inspiring techniques were Spectral subtraction and Spectral gating with Butterworth low pass filter. Experiments results show that the Spectral gating methods gives the highest SNR value of 41.19dB with Butterworth low pass filter.