Could not download file: This paper is available to authorised users only.


A. Hanoosh, I. Marin

University Politehnica of Bucharest (ROMANIA)
During pregnancy can appear complications which have to be closely managed because they endanger the health state of mother and fetus. Panic and stress influence the health state of pregnant women. The medical students and graduates who have their residency program in obstetrics and gynecology are involved in the treatment process. The interpretation is done by them to eliminate the possible pregnancy complications and to decrease the anxiety of their patients.

The scope of the current paper is to present an online 3D nurse that is designed to improve the treatments based on speech recognition and under the assistance of the medical students. The online application monitors remotely the pregnant women who are in the care of the associated medical staff. The non-invasive technique consists in using a multilayer artificial network for analyzing the person's inhale and exhale sounds. The speech recognition uses an image processing convolutional neural network deep learning algorithm and the performance of the NVIDIA graphics processing unit. The frequency coefficients are obtained from the audio recordings and 2D images are created based on them. The model is created as a sequential stack of image layers. The network is trained and the convolutional layer extracts the structural information, after which the evaluation is done. The sounds are classified by the students with the help of the software for determining the panic and the stress levels. Audio instructions are provided by the online 3D nurse according to the detected situation. There has been found that the sound classification is determining the anxiety of the pregnant women. The outcomes help the medical students, obstetrics and gynecology residency graduates to treat their patients.