Hypoxia is a significant condition causing oxygen deficiency in the fetal blood and accounts for more than 23% of perinatal and infant mortality worldwide in a calendar year. Therefore, these circumstances require more efficient meth- ods for prompt detection of hypoxic condition. Cardiotocography (CTG) is the most common technique used to assess fetal well-being and hypoxic complica- tions. Newly, signal processing techniques bring out an innovative horizon for processing the CTG signals. Herein, we are exploring the usefulness of CTG signals by converting them to Recurrence Plots (RP) and classifying using deep learning models for the more accurate detection of hypoxia. A comparative study of VGG16, ResNet and CNN models is done on the RP data. VGG16 achieved better result with an accuracy of 82.02%.