projects

The objective of this project was to develop a deep learning architecture for continuous fetal well-being monitoring by accurately separating fetal ECG signals from abdominal ECG data, even under noisy conditions. To achieve this, we implemented a Residual Convolutional Encoder-Decoder (RCED-Net) model.

The objective of this project was to classify environmental sounds from the UrbanSounds8K dataset into 10 categories - air_conditioner, car_horn, children_playing, dog_bark, drilling, enginge_idling, gun_shot, jackhammer, siren, and street_music. A CNN model was used to classify environmental sounds using either log mel spectrograms or gammatonegrams. Their performances were compared.

This project uses NLP and machine learning to assess semantic similarity between pairs of Quora questions. By preprocessing data, extracting features, and training models, it aims to improve automated question matching and recommendations, enhancing user interactions and accuracy on question-answer platforms.

The objective of this project is to create a new image that follows the content of one image and the style of another image. Neural style transfer technique using a pretrained VGG-19 convolutional neural network was used. The target image was initialized with random noise.

The objective of this project was to uncover the specific auditory features that individual neurons encode by analyzing their spike-triggered averages. This approach enables a deeper understanding of how each neuron's spiking activity relates to particular aspects of a white noise stimulus, revealing functional encoding strategies in the auditory pathway.