|
|
Interpretation of emotions from text on social media
Tlustoš, Vít ; Košař, Vlastimil (oponent) ; Malik, Aamir Saeed (vedoucí práce)
Most human interactions are either text-based or can be converted to text using speech-to-text technologies. This thesis is dedicated to recognizing emotions from these texts. Despite extensive research in this domain, three significant challenges persisted: unexplored or limited cross-domain efficacy of the methods, superficial analysis of the result, and limited usability of the outcomes. We address these challenges by proposing two models based on the RoBERTa model, which we call EmoMosaic-base and EmoMosaic-large. These models were trained on the following datasets: SemEval-2018 Task 1: Affect in Tweets, GoEmotions, XED, and DailyDialog datasets. In contrast to prior studies, we trained our models on all the datasets simultaneously while preserving their original categories. This resulted in models that exhibit strong performance across diverse domains and are directly comparable to other methods. In fact, EmoMosaic-large outperforms recent single-domain state-of-the-art models on SemEval-2018 Task 1: Affect in Tweets and GoEmotions datasets, demonstrating outstanding cross-domain performance. To promote the usability and reproducibility of our research, we make all our code and models public, available at: https://huggingface.co/vtlustos.
|
|
|
Game Development for Assessment of a Person’s Reasoning, Auditory & Visual skills
Pejchar, Štěpán ; Hussain, Yasir (oponent) ; Malik, Aamir Saeed (vedoucí práce)
The goal of the thesis is to implement games that assess cognitive functions of the users playing them, specifically audio-visual cognitive functions and reasoning cognitive functions. The nature of the thesis is a pilot study of a bigger project. The first part of the thesis is the theoretical research. It explains what these cognitive functions are, how they work, and how we can assess them. The second part of the thesis talks about the actual game. It explains the game design and the implementation. The game was implemented in Unity using C\# and a Firebase database. This part also talks about how the game assesses and presents the assessment of the cognitive functions to the users. The last part of the thesis deals with the testing of the game. The game was tested on fifteen users. Their answers are presented and evaluated in the final part.
|
|
|
Implementation of a Multipurpose Measurement System for (Sub)Terahertz Electron Spin Resonance Spectroscopy
Šedivý, Matúš ; Malik, Aamir Saeed (oponent) ; Epel, Boris (oponent) ; Vrba, Radimír (vedoucí práce)
Electron spin resonance (also called electron paramagnetic resonance or just EPR) spectroscopy includes methods that investigate matter via unpaired electrons. One of the progressive EPR methods is the rapid scan, which allows one to observe the kinetics of chemical reactions. Moreover, recent developments of high-frequency components expand application of high-frequency EPR (HFEPR), that use sub terahertz and terahertz waves. This thesis deals with the connection of both paths into the frequency rapid scan (FRaScan) HFEPR spectrometer that was recently developed at CEITEC BUT. A brief introduction to the theoretical background of EPR is provided, followed by an overview of the HFEPR instrumentation. The practical part describes a technical solution for the spectrometer. The emphasis is on the implementation of software through which the spectrometer is controlled, and measurements are automated. Afterward are shown example measurements of solid state materials, namely vanadium doped silicon carbide (SiC:V), lithium phthalocyanine (LiPc), and crystal of 1,3-bisdiphenylene-2-phenylallyl (BDPA). The examples demonstrate the capabilities of the spectrometer to acquire multi-frequency continuous wave spectra and frequency-swept spectra with dependency on the temperature and orientation of the sample, as well as the frequency rapid scan spectra.
|
|
|
Real-Time Processing of Intracranial EEG Signals
Begáň, Patrik ; Malik, Aamir Saeed (oponent) ; Černocký, Jan (vedoucí práce)
In this thesis, we designed and implemented a tool that is able to process intracranial EEG signals in real-time. That is done by applying functions for computing various iEEG biomarkers implemented in python library Epycom on the incoming data stream and storing the results into the database. We compared results computed by our tool against the offline computations and evaluated if real-time signal processing is suitable for clinical practice.
|
|
|
Developing Brain Computer Interface for Imagined Movements
Blašková, Barbora ; Jawed, Soyiba (oponent) ; Malik, Aamir Saeed (vedoucí práce)
Brain disorders and diseases affect 1 in 6 people worldwide and in many cases result in a condition that profusely impacts the life of patient. Mental health topics surge as 1 in 10 people is diagnosed with a mental health disorder. It is therefore crucial to study the organ that is still in a big part a mystery to the researchers - brain. The focus of this thesis is on Brain Computer Interface (BCI) which can act as a intermediary between the brain and a device by acquiring the brain signals and translating them into a set of actions or commands. One of the methods to control a device by thoughts is motor imagery, which is based on the fact that imagining moving a part of the body elicits the same brain response as actual movement. This thesis proposes to utilize a recent field of the EEG for the BCI applications - microstate analysis. Classifier for distinguishing between the motor imagery tasks is proposed as a combination of microstate features extracted from different regions of the brain with the already established features such as from frequency or time-domain. The subject-specific classifiers was trained for 30 participants. Two distinct classifiers were implemented - one for the classification of the rest versus activity and second for the classification of the left versus right motor imagery. The mean accuracy across participants for the rest versus activity classification was 0.85. The mean accuracy across participants for the left versus right motor imagery classification was 0.74. The microstates proved to be helpful in distinguishing between different conditions in a task settings, but need some improvements in terms of the further research.
|
|
|
Emotion Recognition from Brain Electroencephalogram (EEG) Signals
Fritz, Karel ; Jawed, Soyiba (oponent) ; Malik, Aamir Saeed (vedoucí práce)
This study targets classifying emotion states, from Electroencephalogram (EEG) signal. Combining knowledge about physiology of the brain (and emotions), with frequency anal- ysis, complexity analysis, signal processing and deep machine learning (CNN, GNN). Goal of this work is to create the emotion classification model and provide new insights into emotion recognition from EEG. Models created stands on the principles of CNN, GNN, multitask and self supervised training. One of the results achieved State of the Art results on the SEED dataset. Sharing process of understanding this task at the end of the thesis.
|
|
|
Emotion Recognition from Analysis of a Person’s Speech
Knutelský, Martin ; Shakil, Sadia (oponent) ; Malik, Aamir Saeed (vedoucí práce)
This thesis deals with the analysis of emotion recognition from human speech. It aims to design and implement a system that can automatically infer emotional states from speech recordings. The solution is based on the Audio Spectrogram Transformer (AST), a derivative of the Vision Transformer neural network, which accepts mel spectrogram as input. The implementation comprehends the pipeline with two stages. In the first stage, a mel spectrogram is obtained from the input speech recording and in the second stage, the pretrained AST model computes output in the form of probabilities of considered emotional classes. The AST implementation was trained and evaluated on three datasets: RAVDESS, Emo-DB and EMOVO. The obtained results in the form of unweighted accuracy are 84.5 % for RAVDESS, 91.6 % for Emo-DB and 73.8 % for EMOVO. During training, the consumed energy of the graphical processing unit was recorded for the calculation of the carbon footprint in terms of emitted CO2. The main contribution of this work is the utilization of neural network based on Transformer architecture, originally used for vision tasks, to classify emotions from speech. Another contribution is carbon footprint tracking of neural network training. The carbon footprint, expressed in emitted CO2 mass is 1058.37 grams.
|
|
|
Real-Time Processing of Intracranial EEG Signals
Begáň, Patrik ; Malik, Aamir Saeed (oponent) ; Černocký, Jan (vedoucí práce)
In this thesis, we designed and implemented a tool that is able to process intracranial EEG signals in real-time. That is done by applying functions for computing various iEEG biomarkers implemented in python library Epycom on the incoming data stream and storing the results into the database. We compared results computed by our tool against the offline computations and evaluated if real-time signal processing is suitable for clinical practice.
|
host ::
RSS.