Yadisbel Martinez is a PhD researcher at the department of Electronics and Informatics (ETRO) at the Vrije Universiteit Brussel and Universidad de Oriente. She is interested in deep learning models for the Infant cry analysis oriented to the diagnosis and study of infant neurodevelopment.

Research Interests. 

Deep Learning Classification Models for Infant Cry Diagnostic

Infant crying can assume an important role in ensuring the survival, health and development of the child. Through crying, the baby expresses many cues for their basic needs and status thoughts. Consequently, identifying the hidden patterns in the cry signal has been a major challenge. The analysis newborn cries who are suffering from neurological disorders and severe diseases, which can later on result in motor and mental handicap, may prove helpful in early diagnosis of pathologies and protect infants from such disorders. 

"All children – including those with disabilities and developmental delays– need nurturing care and health services to survive and thrive. I hope to help them with my research"

In previous years, many researchers focused on extracting hand-crafted features and used shallow machine learning algorithms for classifying infant cry for diseases diagnoses, such as asphyxia, deaf, autism, hungry, pain and other pathologies. However, in recent years, deep learning has become a highlight in several studies of audio and speech signal processing since they can produce high-level salient features from raw data. In particular, convolutional Mel Frequency Cepstral Coefficients and spectrograms, as inputs in deep learning models for infant cry classification. We will evaluate the performance of both deep learning and traditional machine learning approaches for infant cry classification and demonstrates the power and advantages of deep learning when applied to infant cry signal. Our goal is detect  diseases in Central Nervous System, from the cry signal,  sufficiently in advance for their timely intervention.

  • Yadisbel Martinez is funded from VLIR-UOS Project