Faculteit der Sociale Wetenschappen

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 5 of 1481
  • Item
    Improving the Event-related Potential Classi cation Performance of the Riemannian Pipeline using Shrinkage Regularization
    (2021-06-18) Thoni, Chiara
    The aim in the eld of brain-computer interfaces is to control an electronic device using brain responses that are recorded in an electroencephalogram. These brain signals can be decoded by means of the Riemannian metrics that are associated with the Riemannian manifold. This way of classifying allows for transfer learning possibilities and is more robust opposite to the Euclidean approach. However, to get to these desired characteristics, the data points that represent a single brain response need to be represented by their covariance matrices. This is a drawback of the framework, as the sample covariance matrix is a sub-optimal estimator of the true covariance matrix when the number of samples is low compared to the number of features. To improve the estimator and enhance the classi cation performance, I apply shrinkage regularization on the di erent submatrices of the epoch-based covariance matrix. To assess the e ect of this method, I compare two decoding algorithms against a baseline using six datasets. The results show that there is a signi cant increase in classi cation performance for one out of the six datasets.
  • Item
    Can we go faster than world's fastest brain-computer interface: An application of recurrency to EEG2Code Deep Learning
    (2021-07-01) Teulings, David
    In this research the structure of recurrent neural networks was applied to EEG2Code deep learning. A state of the art high-speed electroencephalogram (EEG) brain-computer interface (BCI) speller, using a deep convolutional neural network. The BCI predicts user intention in a noise-tagging paradigm speller set-up by decoding the EEG data and predicting individual ashes in the noise code. While EEG2Code prides itself as being the current world fasts BCI with an information transfer rate (ITR) of 1237 bits/min with a trial classi cation accuracy of 95:9%. There is still room for improvement on the level of individual ash prediction. Therefore I introduce an extended variant of the EEG2Code deep learning model which makes use of a recurrent neural network layer. The goal of this layer is to capture the temporal dependencies that occur in the data when recording user intention in the hopes of increasing the ability to predict individual ashes with a higher degree of certainty. The recurrent neural network used was a simple recurrent neural network unit. While this addition did not seem to perform as expected, by increasing the performance of individual ash prediction with the initial hyperparameters given, it could serve as a starting point for future researchers to tweak the hyperparameters in an attempt to truly capture the temporal dependencies on an individual ash to ash basis.
  • Item
    A case-based music recommendation system for cochlear implant users
    (2021-05-01) Tanuri Vargas, Andre
    Cochlear implant users face many di culties in their music listening experience, such that music perception and music appreciation remain a challenge. Rhythm is the musical element that is better perceived by CI users when compared to pitch, melody, and timbre. The presence of lyrics and percussive sounds have a positive in uence on the CI users' music appreciation. Music recommendation systems can give appropriate song suggestions to a user. However, the available music recommendation systems are based on timbre and melodic content. In this paper, I propose a case-based music recommendation system that considers music perception and music appreciation from a CI user perceptive.
  • Item
    The impact of sentiment analysis on the user experience of chatbots
    (2021-06-18) Stoutjesdijk, Dylan
    Chatbots have become more popular over the years, however the use of chatbots does not come without problems. This study looked at if sentiment analysis was able to solve these problems and thus see the impact it had on the user experience. The Technology Acceptance Model (TAM) was used to measure the user experience, because it allows an analysis of the impact of an external factor, in this case sentiment analysis, on the user experience. When a user uses a chatbot they have a certain user experience during the interaction. This makes it relevant to measure the experience to see how much sentiment analysis can improve it. Sentiment analysis was used because it is proven that the combination of a chatbot and sentiment seems relevant in making the conversation between a human and a chatbot more clear, sensical and improve the usability (Almansor et al., 2021). The results presented in this experiment showed that sentiment analysis had no impact on the user experience and thus sentiment analysis did not improve the chatbot usability. This could have been because of the small sample size, representation of chatbot users, or even the repetitive side of the chatbot that performed sentiment analysis, or the main topic of the conversation did not require sentiment analysis to be performed.
  • Item
    Music Genre Classi cation Using Gaussian Process Models and Gibbs samplers
    (2021-04-21) Stegge, aan de, Ella
    Music genres are groups with certain musical characteristics. They can be used to create structure within the large amount of music that exists, for example via genre classi cation. Genre classi cation has been automated using Machine Learning. In previous research, Gaussian Processes have successfully been used to classify genres with a reasonable accuracy (60,9% over 6 genres), using relatively few samples (20 per genre). In this paper, several genre classi cation approaches are compared: a Gaussian Process, a Gibbs sampler and as a baseline a Support Vector Ma- chine. First, the optimal kernel for both the Gaussian Process and Support Vector Machine is determined. Then, the models are compared. The models are trained with a range of di erent sample sizes and accuracy is determined using the same testing set. The accuracies that the models achieve are then compared. Using the Support Vector Machine with the polynomial kernel yielded the highest accuracies with the polynomial kernel and the Gaussian Process achieved the highest accuracies with the Mat ern kernel. When trained on the same data, the Gaussian Process outperforms the Support Vector Machine. The Gaussian Process reaches an accuracy of 73,1%, while the Support Vector Machine reaches an accuracy of 61,7%. The Gibbs sampler proves to perform relatively poor on this classi cation problem, as it yields an accuracy of 48,2%.