ABSTRACT
This paper presents GAVT, a highly accurate audiovisual 3D tracking system based on particle filters and a probabilistic framework, employing a single camera and a microphone array. Our first contribution is a complex visual appearance model that accurately locates the speaker's mouth. It transforms a Viola & Jones face detector classifier kernel into a likelihood estimator, leveraging knowledge from multiple classifiers trained for different face poses. Additionally, we propose a mechanism to handle occlusions based on the new likelihood's dispersion. The audio localization proposal utilizes a probabilistic steered response power, representing cross-correlation functions as Gaussian mixture models. Moreover, to prevent tracker interference, we introduce a novel mechanism for associating Gaussians with speakers. The evaluation is carried out using the AV16.3 and CAV3D databases for Single- and Multiple-Object Tracking tasks (SOT and MOT, respectively). GAVT significantly improves the localization performance over audio-only and video-only modalities, with up to 50.3% average relative improvement in 3D when compared with the video-only modality. When compared to the state of the art, our audiovisual system achieves up to 69.7% average relative improvement for the SOT and MOT tasks in the AV16.3 dataset (2D comparison), and up to 18.1% average relative improvement in the MOT task for the CAV3D dataset (3D comparison).
ABSTRACT
We present a novel near-infrared spectroscopy technique based on Dual-Comb optical interrogation (DC-NIRS) applied to dispersive media. The technique recovers the frequency response of the medium under investigation by sampling its spectral response in amplitude and phase. The DC-NIRS reference and sample signals are generated using electro-optic modulation which offers a cost-effective, integrable solution while providing high adaptability to the interrogated medium. A careful choice of both line spacing and optical span of the frequency comb ensures that the retrieved information enables the reconstruction of the temporal impulse response of the medium, known as the diffuse-time-of-flight (DTOF), to obtain its optical properties with a 70 µs temporal resolution and 32 ps photon propagation delay resolution. Furthermore, the DC-NIRS technique also offers enhanced penetration due to noiseless optical amplification (interferometric detection). The presented technique was demonstrated on a static bio-mimetic phantom of known optical properties reproducing a typical brain's optical response. The DTOF and optical properties of the phantom were measured, showing the capabilities of this new technique on the estimation of absolute optical properties with a deviation under 3%. Compared to current technologies, our DC-NIRS technique provides enhanced temporal resolution, spatial location capabilities, and penetration depth, with an integrable and configurable cost-effective architecture, paving the way to next-generation, non-invasive and portable systems for functional brain imaging, and brain-computer interfaces, among other. The system is patent pending PCT/ES2022/070176.
ABSTRACT
Objetivo: Proponer un método basado en el algoritmo de descomposición en modos empíricos para la corrección de los artefactos oculares y cardiacos presentes en el electroencefalograma (EEG).Métodos: Para la corrección de artefactos fue aplicada la reconstrucción parcial dela señal, descartando todos los componentes que pudieran contener información de los artefactos. Luego el método de corrección propuesto es evaluado usando señales de EEG contaminadas artificialmente. Se calcularon los criterios de similitud y corrección entre las salidas del método y el EEG original. Finalmente el método de corrección propuesto fue incorporado a un sistema de monitoreo de anestesia. Para evaluar la mejoría de los resultados del sistema de monitoreo las salidas de este fueron comparadas antes y después de aplicar la corrección de artefactos.Resultados: Las salidas del método de corrección de artefactos en el EEG conservaron una correlación de un 89,7 por ciento entre todas las señales analizadas y un valor medio de similitud de 0,75 comparados con el EEG original.El sistema de monitoreo de anestesia mostró una mejoría de 27,4 por ciento después de aplicar la corrección de losartefactos. Demostrando, el superior desempeño del método de monitoreo de profundidad anestésica propuesto después de realizar la corrección de artefactos.Conclusiones: El EEG se ha convertido en uno de los métodos más usados en la práctica quirúrgica para cuantificar la profundidad anestésica. Pero la exactitud del diagnóstico realizado a partir de esta señal puede verse comprometido por la aparición de artefactos en el registro de EEG. Los artefactos oculares y cardiacos son los más frecuentes y problemáticos(AU)
Objective: To propose a method based on empirical mode decomposition (EMD) algorithm for the correction of eye and cardiac artifacts presents in the electroencephalogram (EEG).Methods: For the artifact correction partial reconstruction of signal were apply the discarding all those components that may contain artifact information. After the proposed correction method is evaluated using artificially contaminated EEG signals. Similitude and correlation criteria were applied between the method outcomesand the original EEG. Finally correction method was incorporated into an anesthesia monitoring system. To evaluate the system outcomes enhancement, these were compares before and after apply the artifact correction.Results: EEG artifact correction method outcomes preserve overall analyzed records a correlation of 89.7 per cent and medium similitude value of 0.75 compared the original EEG. The anesthesia monitoring system shows an enhancement of 27.4 per cent after apply artifact correction.Demonstrating, the superior performance of the anesthetic monitoring proposed methods after artifacts correction.Conclusions: The EEG has become one of the most used method in the surgical practice for to quantify the anesthetic depth. But the accuracy of diagnosis made from this signal can be compromised by the appearance of artifacts in the EEG record. Ocular and cardiac artifacts are most frequent and problematic(AU)
