A parallel probabilistic neural network ECG recognition architecture over GPU platforms
| Autor: | Chakchai So-In, Warintorn Phusomsai, Comdet Phaudphut |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
Discrete wavelet transform
business.industry Time delay neural network Computer science 0206 medical engineering Graphics processing unit Pattern recognition 02 engineering and technology computer.software_genre 020601 biomedical engineering Software framework Support vector machine Probabilistic neural network Autoregressive model 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Artificial intelligence business computer Extreme learning machine |
| Zdroj: | 2016 13th International Joint Conference on Computer Science and Software Engineering (JCSSE). |
| DOI: | 10.1109/jcsse.2016.7748842 |
| Popis: | This paper proposes a novel methodology for ECG recognition in two aspects. The first method is to enhance the recognition rate using Probabilistic Neural Network (PNN) based on its key characteristic of high precision without no-training required as other learning models. Some of its key features are introduced, such as Discrete Wavelet Transform, Autoregressive, and Statistic including some features from the signals after noise mitigation process (e.g., Low and High Pass Filters). The second method is used to further improve the computational time of PNN by investigating the possibility to utilize parallelism based on the advance of Graphics Processing Unit providing an accessible programming framework, such as Compute Unified Device Architecture in a multi-core processing model, and these are Parallel PNN (P-PNN). With a standard MIT-BIH arrhythmia database, the performance of P-PNN is confirmed against well-known learning models, such as Support Vector Machine an Extreme Learning Machine, especially in recognition precision, i.e., around 98% vs. the only 84% and 94%, respectively. Compared with the traditional PNN, P-PNN has also improved the speed-up by factor of four. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|