Ristea Nicolae-Catalin

In this project, I proposed a novel approach for estimating epicentral distance, depth, and magnitude directly from individual raw 3-component seismograms of 1-minute length observed by single stations. The proposed convolutional neural network-based method is able to handle complex-valued representations of the seismic data in the time-frequency domain by using dedicated convolutional and activation functions. The validation experiments were conducted over a publicly available and large database, STanford EArthquake Dataset (STEAD). This is part of a research paper published at IEEE Geoscience and Remote Sensing Letters, a top-tier journal in the geoscience domain.

A data set generator for radar interference mitigation. This is a solution to the lack of publicly available data sets. I proposed a solution based on MATLAB and Python, which generates a custom number of data samples, which mimic real radar data. This project could be used to train deep learning models as well as classical algorithms. This is part of two research papers that were published at VTC-Fall 2020 and CVPR Workshop 2021.

The task of detecting whether a person wears a face mask while speaking is useful in modeling speech in forensic investigations, communication between surgeons, or people protecting themselves against infectious diseases such as COVID-19. In this project, I developed a system for mask detection from speech. My approach is based on training Generative Adversarial Networks (GANs) with cycle-consistency loss to translate unpaired utterances between two classes (with mask and without mask), and on generating new training utterances using the cycle-consistent GANs, assigning opposite labels to each translated utterance. Original and translated utterances are converted into spectrograms which are provided as input to a set of ResNet neural networks with various depths. The networks are combined into an ensemble through a Support Vector Machines (SVM) classifier. With this system, I participated in the Mask Sub-Challenge (MSC) of the INTERSPEECH 2020 Computational Paralinguistics Challenge, surpassing the baseline proposed by the organizers by 2.8%.

I worked on an international project with a public hospital where I developed a system for style transfer between different types of CT (arterial, venous, native) based on Cycle-Transformer-GANs. Moreover, I implemented a method for classification and geometric image registration based on convolutional neural networks (CNNs).