Estudo do potencial evocado visual em regime permanente baseado em LED para interface cérebro máquina

• Ano de defesa: 2011
• Autor(a) principal: Marcos Antonio da Silva Pinto
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: coerência múltipla; Interface cérebro máquina; medida de sincronismo de fase; estimulador visual; potencial evocado; teste espectral F; visual; eletroencefalograma; magnitude quadrática da coerência; Engenharia elétrica
• Link de acesso: http://hdl.handle.net/1843/BUOS-8R3HN7

The Steady State Visual Evoked Potential (SSVEP) is the brains electrical response to intermittent visual stimulation. This has applications in medical clinic and in studies of the neurophysiology of vision and more recently in Brain Machine Interface (BMI). The challenge of the use of the SSVEP is its Low Signal to Noise ratio (SNR) which difficult its identification. Some researchers have shown that in addition to signal processing techniques, improved stimulation techniques can also increase the SNR SSVEP. This work proposes to study techniques of stimulation by LED (Light Emitting Diode) with objective response detection techniques in the frequency domain for fast and high probability detection of SSVEP. A visual stimulator, digital portable LEDbased, able to generate stable and accurate stimuli to SSVEP for BMI applications was developed. Some following stimulation parameters were evaluated: frequency, intensity, color and size. The results showed the highest RSR SSVEP for frequencies of stimulation from 5 to 9 and 25 to 30 Hz, intensities around 15 cd/m², white color and size of 2.86 ° of visual angle. Four objective detection techniques of SSVEP (Spectral F Test SFT, Phase Synchrony Measure PSM, Magnitude Squared of Coherence MSC and Multiple Magnitude Squared of Coherence MMSC) were evaluated. The results indicated that the detector MMSC is the best for applications in BMI for having the highest rates (around 90%) and lowest detection times (about 2 s) for most subjects. These findings may help to make faster and feasible BMIs for everyday use