Infant Cry Audio Recordings From First Day Of Life

6.5 hours of pure expiration sounds from 786 infants, compiled into 26,000 audio files. The dataset was created by Ubenwa and released in 2023, with recordings taken either within the first hour(s) of life or upon hospital discharge within 24 hours of birth.

Use Cases

• Classify infant cries by recording context using the provided 'time of recording' feature (first hour vs. discharge).
• Train models for audio-based health assessment using the 26k audio files of pure expiration sounds.
• Analyze acoustic patterns in cries from 786 unique infant subjects for biomarker discovery.
• Benchmark audio classification algorithms on the CryCeleb2023 competition task.

Strengths

• 26,000 audio files providing substantial data volume for model training
• 6.5 hours of curated, pure expiration sounds
• Sourced from 786 infant subjects across multiple hospitals

Limitations

• Limited temporal scope, with all recordings confined to the first 24 hours after birth
• Unknown class distribution or balance across subjects or recording conditions
• No information provided on audio quality metrics, sample rates, or formats

Provenance

Source

Ubenwa via Hugging Face

Collection Method

Cries gathered from infants in various hospitals.

Time Range

Recordings made within the first hour(s) of life or at hospital discharge within 24 hours of birth.

Freshness

Last updated in October 2023.

Geography

Multiple hospitals, specific locations unknown.