Normal and hypoacoustic infant cry signal classification using time-frequency analysis and general regression neural...

Area of Science:

• Biomedical Engineering
• Acoustic Signal Processing
• Infant Health

Background:

• Infant crying is a primary communication method.
• Acoustic analysis of cries can reveal infant health status.
• Pathological cries may indicate underlying conditions, such as deafness.

Purpose of the Study:

• To propose a novel method for analyzing infant cry signals using time-frequency analysis.
• To differentiate between normal and pathological infant cries, specifically from deaf infants.
• To evaluate the effectiveness of different machine learning classifiers for this task.

Main Methods:

• Short-Time Fourier Transform (STFT) for time-frequency analysis of infant cry signals.
• Extraction of statistical features from the time-frequency plots.
• Classification using General Regression Neural Network (GRNN), Multilayer Perceptron (MLP), and Time-Delay Neural Network (TDNN).

Main Results:

• The GRNN classifier demonstrated superior performance in distinguishing between normal and pathological infant cries.
• The proposed feature extraction method combined with GRNN achieved high classification accuracy.
• Comparison with MLP and TDNN confirmed the effectiveness of the GRNN approach.

Conclusions:

• Acoustic analysis of infant cries, particularly using STFT and statistical features, is a viable tool for identifying pathological cries.
• GRNN provides a reliable and accurate method for classifying infant cries, aiding in the early detection of conditions like deafness.
• The proposed methodology offers a promising non-invasive approach to infant health assessment.