Advances in Infant Cry Paralinguistic Classification-Methods, Implementation, and Applications: Systematic Review.

Area of Science:

• Signal Processing
• Machine Learning
• Infant Health

Background:

• Infant cries are crucial for communication, but traditional interpretation is subjective and slow.
• Precise cry analysis can offer vital insights into infant health and needs.
• Timely and accurate responses are essential for infant well-being and caregiving.

Purpose of the Study:

• To systematically review advancements in infant cry classification over 24 years.
• To analyze methods, coverage, deployment, and applications of infant cry analysis.
• To identify trends and future directions in infant cry signal processing.

Main Methods:

• Conducted a systematic literature review across 9 electronic databases.
• Screened 5904 results, selecting 126 eligible studies.
• Assessed study quality using Cochrane RoB2 and TRIPOD guidelines.

Main Results:

• Machine learning, deep learning, and hybrid models have advanced infant cry classification since 2019.
• Common features include Mel-frequency cepstral coefficients and spectrograms.
• Most models remain undeployed; limited use of denoising and federated learning.

Conclusions:

• Infant cry classification evolved to ML but lacks focus on privacy and deployment.
• Future research should standardize multimodal approaches and incorporate federated learning for data confidentiality.
• Proposed denoising layer and broader audio features will improve model accuracy and applicability.