In this paper a novel speaker recognition system is introduced. Automated speaker recognition has become increasingly popular to aid in crime investigations and authorization processes with the advances in computer science. Here, a recurrent neural network approach is used to learn to identify ten speakers within a set of 21 audio books. Audio signals are processed via spectral analysis into Mel Frequency Cepstral Coefficients that serve as speaker specific features, which are input to the neural network. The Long Short-Term Memory algorithm is examined for the first time within this area, with interesting results. Experiments are made as to find the optimum network model for the problem. These show that the network learns to identify the speakers well, text-independently, when the recording situation is the same. However the system has problems to recognize speakers from different recordings, which is probably due to noise sensitivity of the speech processing algorithm in use.

Made available in DSpace on 2010-04-20T21:00:53Z (GMT). No. of bitstreams: 4 Eduardo Menescal Lustosa Longo.pdf.jpg: 10145 bytes, checksum: b065ece0a6050c28545e0c2004a42402 (MD5) Eduardo Menescal Lustosa Longo.pdf.txt: 40329 bytes, checksum: 1026ba29309ee769eab96eae93bebd92 (MD5) Eduardo Menescal Lustosa Longo.pdf: 273765 bytes, checksum: c3f0204a9f073bb4276db2a68130c0cb (MD5) license.txt: 4886 bytes, checksum: a7c9cfe8632065c9656e1d827d849400 (MD5) Previous issue date: 2009-01-29T00:00:00Z / In this dissertation, we assess the application of pairs trading strategy in the Brazilian stock market. Differently from other papers about the same theme, we built synthetic asset from a linear relationship between stock prices. Accordingly to Burgeois and Minko (2005), we applied the Johansen methodology to identify stock pairs to be evaluated. After identifying cointegrated stock pairs, in order to filter out nonstationary synthetic assets, we did the DF-GLS and KPSS tests and removed those with unit root in the time series. Thereafter, we simulate the strategy (backtesting) in the selected pairs and in order to find the best results parameters, we optimized the results using different formation periods, trading periods and entry, exit and stop-loss parameters. To assure the most close to reality backtesting, we included in the results the brokerage, exchange and stock loan fees. In addition we added one day lag to trade after an order has triggered. / Neste trabalho, verificamos viabilidade de aplicação da estratégia de pairs trading no mercado acionário brasileiro. Diferentemente de outros estudos do mesmo tema, construímos ativos sintéticos a partir de uma combinação linear de preços de ações. Conforme Burgeois e Minko (2005), utilizamos a metodologia de Johansen para a formação dos pares a serem testados. Após a identificação de pares cointegrados, para assegurar a estacionaridade do ativo sintético contruído a partir da relação linear de preços das ações, utilizamos os testes DF-GLS e KPSS e filtramos àqueles que apresentavam raiz unitária em sua série de tempo. A seguir, simulamos a estratégia (backtesting) com os pares selecionados e para encontrar os melhores parâmetros, testamos diferentes períodos de formação dos pares, de operação e de parâmetros de entrada, saída e stop-loss. A fim de realizarmos os testes de forma mais realista possível, incluímos os custos de corretagem, de emolumentos e de aluguel, além de adicionar um lag de um dia para a realização das operações

Submitted by Ronaldo Cesar Dametto (rdametto@uol.com.br) on 2018-09-18T19:17:34Z No. of bitstreams: 1 Dissertação_Completa_Final.pdf: 2885777 bytes, checksum: 05b2d5417efbec72f927cf8a62eef3fb (MD5) / Approved for entry into archive by Lucilene Cordeiro da Silva Messias null (lubiblio@bauru.unesp.br) on 2018-09-20T12:19:07Z (GMT) No. of bitstreams: 1 dametto_rc_me_bauru.pdf: 2877027 bytes, checksum: cee33d724090a01372e1292109af2ce9 (MD5) / Made available in DSpace on 2018-09-20T12:19:07Z (GMT). No. of bitstreams: 1 dametto_rc_me_bauru.pdf: 2877027 bytes, checksum: cee33d724090a01372e1292109af2ce9 (MD5) Previous issue date: 2018-08-06 / O aprendizado de máquina vem sendo utilizado em diferentes segmentos da área financeira, como na previsão de preços de ações, mercado de câmbio, índices de mercado e composição de carteira de investimento. Este trabalho busca comparar e combinar três tipos de algoritmos de aprendizagem de máquina, mais especificamente, o método Ensemble de Redes Neurais Artificias com as redes Multilayer Perceptrons (MLP), auto-regressiva com entradas exógenas (NARX) e Long Short-Term Memory (LSTM) para predição do Índice Bovespa. A amostra da série do Ibovespa foi obtida pelo Yahoo!Finance no período de 04 de janeiro de 2010 a 28 de dezembro de 2017, de periodicidade diária. Foram utilizadas as séries temporais referentes a cotação do Dólar, além de indicadores numéricos da Análise Técnica como variáveis independentes para compor a predição. Os algoritmos foram desenvolvidos através da linguagem Python usando framework Keras. Para avaliação dos algoritmos foram utilizadas as métricas de desempenho MSE, RMSE e MAPE, além da comparação entre as previsões obtidas e os valores reais. Os resultados das métricas indicam bom desempenho de predição pelo modelo Ensemble proposto, obtendo 70% de acerto no movimento do índice, porém, não conseguiu atingir melhores resultados que as redes MLP e NARX, ambas com 80% de acerto. / Different segments of the financial area, such as the forecast of stock prices, the foreign exchange market, the market indices and the composition of investment portfolio, use machine learning. This work aims to compare and combine two types of machine learning algorithms, the Artificial Neural Network Ensemble method with Multilayer Perceptrons (MLP), auto-regressive with exogenous inputs (NARX) and Long Short-Term Memory (LSTM) for prediction of the Bovespa Index. The Bovespa time series samples were obtained daily, using Yahoo! Finance, from January 4th, 2010 to December 28th, 2017. Dollar quotation, Google trends and numerical indicators of the Technical Analysis were used as independent variables to compose the prediction. The algorithms were developed using Python and Keras framework. Finally, in order to evaluate the algorithms, the MSE, RMSE and MAPE performance metrics, as well as the comparison between the obtained predictions and the actual values, were used. The results of the metrics indicate good prediction performance by the proposed Ensemble model, obtaining a 70% accuracy in the index movement, but failed to achieve better results than the MLP and NARX networks, both with 80% accuracy.

Natural Language Processing (NLP) is a field studying computer processing of human language. Recently, neural network language models, a subset of machine learning, have been used to great effect in this field. However, research remains focused on the English language, with few implementations in other languages of the world. This work focuses on how NLP techniques can be used for the task of grammar and spelling correction in the Swedish language, in order to investigate how language models can be applied to non-English languages. We use a controlled experiment to find the hyperparameters most suitable for grammar and spelling correction on the Göteborgs-Posten corpus, using a Long Short-term Memory Recurrent Neural Network. We present promising results for Swedish-specific grammar correction tasks using this kind of neural network; specifically, our network has a high accuracy in completing these tasks, though the accuracy achieved for language-independent typos remains low.

When moving into a more connected world together with machines, a mutual understanding will be very important. With the increased availability in wear- able sensors, a better understanding of human needs is suggested. The Dart- mouth Research study at the Psychiatric Research Center has examined the viability of detecting and further on predicting human behaviour and complex tasks. The field of smoking detection was challenged by using the Q-sensor by Affectiva as a prototype. Further more, this study implemented a framework for future research on the basis for developing a low cost, connected, device with Thayer Engineering School at Dartmouth College. With 3 days of data from 10 subjects smoking sessions was detected with just under 90% accuracy using the Conditional Random Field algorithm. However, predicting smoking with Electrodermal Momentary Assessment (EMA) remains an unanswered ques- tion. Hopefully a tool has been provided as a platform for better understanding of habits and behaviour.

Given the volatility of the stock market and the multitude of financial variables at play, forecasting the value of stocks can be a challenging task. Nonetheless, such prediction task presents a fascinating problem to solve using machine learning. The stock market can be affected by news events, social media posts, political changes, investor emotions, and the general economy among other factors. Predicting the stock value of a company by simply using financial stock data of its price may be insufficient to give an accurate prediction. Investors often openly express their attitudes towards various stocks on social medial platforms. Hence, combining sentiment analysis from social media and the financial stock value of a company may yield more accurate predictions. This thesis proposes a method to predict the stock market using sentiment analysis and financial stock data. To estimate the sentiment in social media posts, we use an ensemble-based model that leverages Multi-Layer Perceptron (MLP), Long Short-Term Memory (LSTM), and Convolutional Neural Network (CNN) models. We use an LSTM model for the financial stock prediction. The models are trained on the AAPL, CSCO, IBM, and MSFT stocks, utilizing a combination of the financial stock data and sentiment extracted from social media posts on Twitter between the years 2015-2019. Our experimental results show that the combination of the financial and sentiment information can improve the stock market prediction performance. The proposed solution has achieved a prediction performance of 74.3%.

The thesis focuses on the use of deep recurrent neural network, architecture Long Short-Term Memory for robust denoising of audio signal. LSTM is currently very attractive due to its characteristics to remember previous weights, or edit them not only according to the used algorithms, but also by examining changes in neighboring cells. The work describes the selection of the initial dataset and used noise along with the creation of optimal test data. For network training, the KERAS framework for Python is selected. Candidate networks for possible solutions are explored and described, followed by several experiments to determine the true behavior of the neural network.

In industry, it is crucial to be able to detect damage or abnormal behavior in machines. A machine's downtime can be minimized by detecting and repairing faulty components of the machine as early as possible. It is, however, economically inefficient and labor-intensive to detect machine fault sounds manual. In comparison with manual machine failure detection, automatic failure detection systems can reduce operating and personnel costs.  Although prior research has identified many methods to detect failures in drill machines using vibration or sound signals, this field still remains many challenges. Most previous research using machine learning techniques has been based on features that are extracted manually from the raw sound signals and classified using conventional classifiers (SVM, Gaussian mixture model, etc.). However, manual extraction and selection of features may be tedious for researchers, and their choices may be biased because it is difficult to identify which features are good and contain an essential description of sounds for classification. Recent studies have used LSTM, end-to-end 1D CNN, and 2D CNN as classifiers for classification, but these have limited accuracy for machine failure detection. Besides, machine failure occurs very rarely in the data. Moreover, the sounds in the real-world dataset have complex waveforms and usually are a combination of noise and sound presented at the same time. Given that drill failure detection is essential to apply in the industry to detect failures in machines, I felt compelled to propose a system that can detect anomalies in the drill machine effectively, especially for a small dataset. This thesis proposed modern artificial intelligence methods for the detection of drill failures using drill sounds provided by Valmet AB. Instead of using raw sound signals, the image representations of sound signals (Mel spectrograms and log-Mel spectrograms) were used as the input of my proposed models. For feature extraction, I proposed using deep learning 2-D convolutional neural networks (2D-CNN) to extract features from image representations of sound signals. To classify three classes in the dataset from Valmet AB (anomalous sounds, normal sounds, and irrelevant sounds), I proposed either using conventional machine learning classifiers (KNN, SVM, and linear discriminant) or a recurrent neural network (long short-term memory). For using conventional machine learning methods as classifiers, pre-trained VGG19 was used to extract features and neighborhood component analysis (NCA) as the feature selection. For using long short-term memory (LSTM), a small 2D-CNN was proposed to extract features and used an attention layer after LSTM to focus on the anomaly of the sound when the drill changes from normal to the broken state. Thus, my findings will allow readers to detect anomalies in drill machines better and develop a more cost-effective system that can be conducted well on a small dataset. There is always background noise and acoustic noise in sounds, which affect the accuracy of the classification system. My hypothesis was that noise suppression methods would improve the sound classification application's accuracy. The result of my research is a sound separation method using short-time Fourier transform (STFT) frames with overlapped content. Unlike traditional STFT conversion, in which every sound is converted into one image, a different approach is taken. In contrast, splitting the signal into many STFT frames can improve the accuracy of model prediction by increasing the variability of the data. Images of these frames separated into clean and noisy ones are saved as images, and subsequently fed into a pre-trained CNN for classification. This enables the classifier to become robust to noise. The FSDNoisy18k dataset is chosen in order to demonstrate the efficiency of the proposed method. In experiments using the proposed approach, 94.14 percent of 21 classes were classified successfully, including 20 classes of sound events and a noisy class. / <p>Vid tidpunkten för disputationen var följande delarbeten opublicerade: delarbete 2 och 3 inskickat.</p><p>At the time of the doctoral defence the following papers were unpublished: paper 2 and 3 submitted.</p> / AISound – Akustisk sensoruppsättning för AI-övervakningssystem / MiLo — miljön i kontrolloopen

The thesis focuses on the use of deep recurrent neural network, architecture Long Short-Term Memory for robust denoising of audio signal. LSTM is currently very attractive due to its characteristics to remember previous weights, or edit them not only according to the used algorithms, but also by examining changes in neighboring cells. The work describes the selection of the initial dataset and used noise along with the creation of optimal test data. For creation of the training network is selected KERAS framework for Python and are explored and discussed possible candidates for viable solutions.

In this thesis, methods of implementing an online gesture recognition system for the novel Stylaero Board device are investigated. Two methods are evaluated - one based on LSTMs and one based on HMMs - on three kinds of gestures: Tap, circle, and flick motions. A method’s performance was measured in its accuracy in determining both whether any of the above listed gestures were performed and, if so, which gesture, in an online single-pass scenario. Insight was acquired regarding the technical challenges and possible solutions to the online aspect of the problem. Poor performance was, however, observed in both methods, with a likely culprit identified as low quality of training data, due to an arduous and complex gesture performance capturing process. Further research improving on the process of gathering data is suggested.