Objective Child Vocal Development Measurement with Naturalistic Daylong Audio Recording

Child vocal development is a subject that touches many areas. Its measurement is based mainly on subjective approaches. This study demonstrates an objective and unobtrusive measurement and monitoring approach using daylong audio recordings of the natural home environment. One purpose of this study is to explore the underlying features about child vocal development buried within the audio streams of daylong recordings using automated approaches with signal processing and pattern recognition technologies. Various child vocal development features are studied, including phonetic development features, features based on unsupervised and self-organized sound categories, features about phone sequence information, prosodic features, spectrum features and so on. We are interested in the developmental trends and the group differences of the features. Based on the developmental trends of the features, the vocalization age can be estimated. A correlation of 0.84 between the estimated vocalization age and the chronological age for children of typical development is achieved. Based on the group differences, the features can be used for childhood autism identification. 94% accuracy at equal-error rate point is achieved for this purpose. Similar to many emerging non-invasive and telemonitoring technologies in health care, this approach is believed to have great potential in child development research, clinical practice and parenting.