Research
Here some of my research I did during my master.
Classification of pneumonia images using CCN
Abstract : Classification of pneumonia images using Convolutional neural network to indentify normal lung with or withour pneumonia.
Agent-based model of measurement of personal proteccion against the spread of the SARS-COVID-2
Abstract : After having presented several waves of spread of the COVID-19 (SARS-COV 2) pandemic, all countries of the world are struggling to regain their economy by slowly lifting restrictions on mobility and social distancing measures applied during the crisis. Numerous studies provide us with evidence convincing about how contact distancing, face mask use, and hand washing habits can reduce the risk of transmission of SARS-CoV-2. This work shows the research carried out on Agent-based models to simulate the spread of COVID-19 incorporating personal protection measures.
Speech recognition in real time for advertisements in audio streaming
Abstract : This research describes a project to classify audio signals captured from a streaming in real time. To carry out the signal processing, a file of approximately eight hours duration was obtained, which was divided into 3-second intervals, allowing to obtain the characteristics of the selected advertisements (slogans). A part of these files will be used as training data of the selected model and the other part for its validation. 28 attributes were initially selected using the spectrogram of the signal. Likewise, a classification neural network for multistate labels is modeled using the MLP Classifier algorithm and applying the softmax output activation function in the last layer.
Identify the age of speaker by Speech recognition
Abstract : One of the problems in automatic speaker recognition systems is voice changes due to factors environmental such as external noises, from vocal effort when raising or lowering the voice intentionally or hoarseness or when using unusual speech accents that correspond to specific regions or also the age of the people. This last point is analyzed in this work with which it is intended to find those characteristics that allow identifying the age of the people from their identified voice.
Optimize 3D bin packing problem with PSO and GRASP metaheuristic
Abstract : We present the problem of distribution of boxes (items) in containers (bins) which is based on the “3D bin packing problem” problem. In the present work we study 2 metaheuristics PSO and GRASP studied in class and 1 heuristic related to packing boxes within a container called EMS (Empty Maximal Spaces). We also include results and conclusions from the work done.
News text clasification based in its content as supervised and unsupervised Clustering LDA
Abstract : The intention of this work is to identify the classification of news by predicting membership group using a previously supervised classification model trained using Recurrent Networks like LSTM and GRU and compare the results obtained with the application of an LDA clustering model.