Loading...
7 results
Search Results
Now showing 1 - 7 of 7
- Métodos de seleção de parâmetros para o diagnóstico de patologias da laringePublication . Silva, Letícia; Teixeira, João Paulo; Bispo, Bruno CatarinoEsta dissertação propõe soluções para a identificação de patologias da voz através do processamento do sinal de fala. Foram utilizados na classificação de patologias como Laringite Crónica, Disfonia e Paralisia das Cordas Vocais as redes neuronais, Multilayer Perceptron e Long-Short-Term-Memory. Os parâmetros acústicos empregados foram jitter relativo, jitter absoluto, shimmer relativo, shimmer absoluto, autocorrelação, Harmonic to Noise Ratio, Noise to Harmonic Ratio e Mel Frequency Cepstral Coefficients. Estes parêmetros são extraídos da base de dados Saarbrücken Voice Database, a partir de arquivos de áudio que contém as vogais sustentadas /a/, /i/ e /u/ nos tons baixo, normal e alto. Também empregou e testou técnicas de normalização de dados, identificação de outliers e seleção de parâmetros. Tais aplicações tem a finalidade de otimizar o modelo de reconhecimento, torná-lo mais eficiente e consequentemente melhorar a acurácia/exatidão do diagnóstico. Como pré-processamento utilizou-se as técnicas de normalização Z-score, Logarítmica e Raiz Quadrada para permitir uma melhor identificação dos outliers presente nos dados, por meio da aplicação do método do Box Plot e do Desvio Padrão. Após os experimentos, tanto o método do Desvio Padrão quanto o do Box Plot com normalização do Z-score mostraram-se muito úteis para o pré-processamento do conjunto de dados para o reconhecimento de patologias de voz. A acurácia foi melhorada entre 3 a 13 pontos em percentagem. Posteriormente, foram utilizadas as técnicas de Seleção de Parâmetros que ordenam os atributos segundo uma métrica de importância. Deste modo, os parâmetros relevantes são selecionados de acordo com o critério estabelecido pelos testes: Correlação, ReliefF, Test t de Welch, Regressão Multilinear. Ao comparar todos os algoritmos desenvolvidos, pode-se destacar que o algoritmo baseado no ReliefF teve o melhor desempenho. Com relação a acurácia teve um aumento de 9 pontos percentuais e na medida F de 8 pontos percentuais.
- The digital twin paradigm applied to soil quality assessment: a systematic literature reviewPublication . Silva, Letícia; Rodríguez-Sedano, Francisco Jesús; Baptista, Paula; Coelho, João PauloThis article presents the results regarding a systematic literature review procedure on digital twins applied to precision agriculture. In particular, research and development activities aimed at the use of digital twins, in the context of predictive control, with the purpose of improving soil quality. This study was carried out through an exhaustive search of scientific literature on five different databases. A total of 158 articles were extracted as a result of this search. After a first screening process, only 11 articles were considered to be aligned with the current topic. Subsequently, these articles were categorised to extract all relevant information, using the preferred reporting items for systematic reviews and meta-analyses methods. Based on the obtained results, there are two main conclusions to draw: First, when compared with industrial processes, there is only a very slight rising trend regarding the use of digital twins in agriculture. Second, within the time frame in which this work was carried out, it was not possible to find any published paper on the use of digital twins for soil quality improvement within a model predictive control context.
- Promoting olive groves’s soil quality by a digital twin’s predictive based control: the sensor’s networkPublication . Silva, Letícia; Giugge, Romina Eliana; Rodríguez-Sedano, Francisco Jesús; Baptista, Paula; Coelho, João PauloIn Portugal, the olive groves soil has been subject to severe degradation due to harsh farming techniques combined with uncontrollable conditions such as climate changes and steep terrain orography. In this framework, this tendency must be reverted by adopting different farm management policies. The MAN4HEALTH project address this subject in two-folds: one, at an agronomic level, where the soil will be protected by growing a layer of indigenous plants and other, that resort to the soil’s digitization in order to improve the deployment of fertilizers. This paper addresses the latter and aims to provide an overall description of the architecture of a predictive control system based on the soil’s digital twin. In this control paradigm, an artificial intelligence layer will follow the entire cultivation process through the development of a soil’s digital twin. This model will be able to describe the spatial and temporal dynamics of fertilization policies and be included within a model predictive control strategy in order to both decrease the concentration of chemicals released into the soil and promote the economic income of the farmer. In particular, this paper tackles one of the phases of this project where soil digitization must be carried out in order to feed the data to the digital twin. In particular, the description of the sensor network and the data management architecture.
- Transfer learning with audioSet to voice pathologies identification in continuous speechPublication . Guedes, Victor; Teixeira, Felipe; Oliveira, Alessa Anjos de; Fernandes, Joana Filipa Teixeira; Silva, Letícia; Candido Junior, Arnaldo; Teixeira, João PauloThe classification of pathological diseases with the implementation of concepts of Deep Learning has been increasing considerably in recent times. Among the works developed there are good results for the classification in sustained speech with vowels, but few related works for the classification in continuous speech. This work uses the German Saarbrücken Voice Database with the phrase “Guten Morgen, wie geht es Ihnen?” to classify four classes: dysphonia, laryngitis, paralysis of vocal cords and healthy voices. Transfer learning concepts were used with the AudioSet database. Two models were developed based on Long-Short-Term-Memory and Convolutional Network for classification of extracted embeddings and comparison of the best results, using cross-validation. The final results allowed to obtaining 40% of f1-score for the four classes, 66% f1-score for Dysphonia x Healthy, 67% for Laryngitis x healthy and 80% for Paralysis x Healthy.
- Outliers treatment to improve the recognition of voice pathologiesPublication . Silva, Letícia; Hermsdorf, Juliana; Guedes, Victor; Teixeira, Felipe; Fernandes, Joana Filipa Teixeira; Bispo, Bruno; Teixeira, João PauloIn some of the processes used in data analysis, such as the recognition of pathologies and pathological subjects, the presence of anomalous instances in the dataset is an unfavorable situation that can lead to misleading results. This article presents a function that implements the identification of anomalies in dataset using the boxplot and standard deviation methods. Also was used the filling technique to treat these anomalies, in which the anomalous point value were substituted by a limit value determined by the boxplot or standard deviation methods. To improve the outliers methods some normalization processes based on the z-score, logarithmic and squared root methodologies were experimented. These outliers treatment were applied to the dataset used in the recognition of vocal pathologies (dysphonia, chronic laryngitis and vocal cords paralysis vs control), performed by a MLP and LSTM neural networks. After the experiments, both the standard deviation and the boxplot methods with z-score normalization showed very useful for pre-processing the dataset for voice pathologies recognition. The accuracy was improved between 3 and 13 points in percentage.
- Features Selection Algorithms for Classification of Voice SignalsPublication . Silva, Letícia; Bispo, Bruno; Teixeira, João PauloIn data mining problems, the high dimensionality of the input features can affect the performance of the process. In this way, the features selection methods appear as a solution to the problems encountered when analyzing databases with large dimensions. This article presents the implementation of the Pearson's linear correlation, ReliefF, Welch's t-test and multilinear regression based algorithms with forwards selection and backward elimination direction for the selection of acoustic features for the task of voice pathologies identification. The best set of selected features improved the accuracy and F1-score from 83% to 92% (9 points of percentage), using the ReliefF algorithm.
- Parameters for vocal acoustic analysis - cured databasePublication . Fernandes, Joana Filipa Teixeira; Silva, Letícia; Teixeira, Felipe; Guedes, Victor; Santos, Juliana Hermsdorf; Teixeira, João PauloThis paper describes the construction and organization of a database of speech parameters extracted from a speech database. This article intends to inform the community about the existence of this database for future research. The database includes parameters extracted from sounds produced by patients distributed among 19 diseases and control subjects. The set of parameters of this database consists of the jitter, shimmer, Harmonic to Noise Ratio (HNR), Noise to Harmonic Ratio (NHR), autocorrelation and Mel Frequency Cepstral Coefficients (MFCC) extracted from the sound of sustained vowels /a/, /i/ and /u/ at the high, low and normal tones, and a short German sentence. The cured database has a total number of 707 pathological subjects (distributed by the various diseases) and 194 control subjects, in a total of 901 subjects.