Navegando por Assunto "Data processing"

Agora exibindo 1 - 3 de 3

Acesso aberto (Open Access)
Aplicabilidade de um sistema de baixo custo para o monitoramento de dados meteoceanográficos na zona costeira Amazônica
(Universidade Federal do Pará, 2025-04-23) SOTÃO, Daniel da Silva; ROSÁRIO, Renan Peixoto; http://lattes.cnpq.br/8003860457518342; https://orcid.org/0000-0003-2913-0514
This study aimed to develop, implement, and validate a low-cost Continuous Monitoring Prototype (CMP) to collect meteoceanographic data in the Amazon Coastal Zone (ACZ), a region vulnerable to climate change and lacking environmental data. The system, built with an ESP32 microcontroller, integrates sensors for air temperature and relative humidity (HDC1080 and AM2302), atmospheric pressure (BMP280 and MS5611), water temperature (DS18B20), tide level (HC-SR04), precipitation, and wind speed and direction. Sensor validation was performed by comparing the collected data with reference equipment: a CICLUS PRO weather station (CWS) and a Sonlist Levelogger 5 LTC recorder (CTDlog). The CMP conducted four joint test campaigns with the CWS and CTDlog, totaling 56,221 records. Linear regression was applied to develop calibration equations for each sensor. Calibration quality was assessed using the coefficient of determination (R²), root mean square error (RMSE), Pearson correlation coefficient, and residual analysis. The HDC1080 and AM2302 sensors showed excellent performance for temperature measurement, with R² > 0.9, RMSE < 0.2 °C, mean absolute residual (MAE) < 0.12 °C, and very strong Pearson correlation (r ≥ 0.9). However, they exhibited instability in relative humidity measurements (R² ≈ 0.64; RMSE ≈ 3.46%; MAE ≈ 2.52%). The performance of the BMP280 and MS5611 pressure sensors was initially affected by internal heating, but after thermal compensation, they achieved R² values between 0.88 and 0.99, RMSE from 0.17 to 0.45 hPa, and MAE from 0.11 to 0.34 hPa, with the BMP280 showing superior performance. The DS18B20 water temperature sensor showed very strong correlation, with R² ≈ 0.94, RMSE ≈ 0.036 °C, and MAE ≈ 0.021 °C, proving highly promising. The HCSR04 sensor stood out as the most accurate, with R² ≈ 0.99, RMSE ≈ 2.6 cm, and MAE ≈ 1.9 cm. Precipitation and wind sensors presented inconsistencies and require further testing. The total production cost of the CMP was BRL 952.75, representing savings of 86.19% compared to the CWS (BRL 6,897.00) and 96.14% compared to the CTDlog (BRL 24,677.29). When compared to equivalent equipment, the CMP achieved at least 66.33% cost reduction compared to basic stations and 87.3% compared to lower-cost temperature and water level recorders. The CMP proved to be a viable, economical, and replicable solution for continuous environmental monitoring in the ACZ, with the potential to fill existing gaps in observation networks, although improvements are still needed to enhance its long-term performance.
Acesso aberto (Open Access)
Joining diffraction filter and residual diffraction moveout to constructo a velocity model in the depth and time domains: application to a Viking Graben data set
(Universidade Federal do Pará, 2014-08-14) COLLAZOS GONZALEZ, Jaime Andres; FIGUEIREDO, José Jadsom Sampaio de; http://lattes.cnpq.br/1610827269025210
Diffracted seismic waves are generated by unsmooth structures in the subsurface with a size on the order of seismic wavelengths. Because the incident wave field can be significantly affected by these discontinuities, many important properties of the seismic events can be used to improve the velocity model building. In this thesis, we propose a practical approach to construct velocity models in the time and depth domains using diffractions. This methodology applies the plane wave destruction (PWD) filter jointly with the residual diffraction moveout (RDM) method to construct velocity models in time and depth domains. Our method does not depend on any requirements except for identifiable diffractions filtered from reflection events and an arbitrary initial velocity model as input. The post-stack migrated images (in the time and depth domains) are compared with the mi- grated images derived from conventional seismic processing steps. In both cases, we used post-stack Kirchhoff Migration. Beyond the to the need to identify and select the diffraction events in the post-stack migrated sections in the depth domain, the method has a very low computational cost of processing time. To reach an acceptable velocity model was less compared with conventional processing. The applicability of our methodology was verified using a real Viking Graben seismic dataset.
Acesso aberto (Open Access)
Um software de reconhecimento de voz para português brasileiro
(Universidade Federal do Pará, 2010-06-17) SILVA, Carlos Patrick Alves da; KLAUTAU JÚNIOR, Aldebaro Barreto da Rocha; http://lattes.cnpq.br/1596629769697284
This work describes a speech recognition software for Brazilian Portuguese. The main objective is to build a system for large vocabulary continuous speech recognition, able to be used in real-time applications. The concepts, characteristics and all steps needed for the construction of such systems are presented. Several resources were produced and made available: acoustic and language models, new voice and text corpora. The text corpus has been built through the extraction and automatic formatting of text from newspapers on the Internet. In addition, two voice corpora were produced, one based on audiobooks and another specifically developed to simulate real-time tests. This work also proposes the use of speaker adaptation techniques for solving the acoustic mismatch problem between speech corpora. Finally, an application programming interface is presented in order to facilitate using the open-source Julius speech decoder. Performance tests are also presented, comparing the developed system with a commercial software.