This paper introduces embedded neural networks for object recognition incorporating parameters such as stiffness, size, and shape. A multisensory glove integrating five force-resistive (FSR) sensors and five inertial measurement (IMU) units per finger was developed to collect data from 16 objects of varying size, shape, and stiffness. The sensor data were processed through three machine learning models namely a Multi-Layer Perceptron (MLP), a Single-Layer (SLP) model, and a Convolutional Neural Network (CNN). The results show that the MLP model exhibited the highest accuracy for the fusion of hard and soft objects, achieving an accuracy of 99.5%. Overall, the proposed system shows promising potential as a wearable feedback system for post-stroke rehabilitation.