Eye and Voice-Controlled Human Machine Interface System for Wheelchairs Using Image Gradient Approach.

Rehabilitative mobility aids are being used extensively for physically impaired people. Efforts are being made to develop human machine interfaces (HMIs), manipulating the biosignals to better control the electromechanical mobility aids, especially the wheelchairs. Creating precise control commands such as move forward, left, right, backward and stop, via biosignals, in an appropriate HMI is the actual challenge, as the people with a high level of disability (quadriplegia and paralysis, etc.) are unable to drive conventional wheelchairs. Therefore, a novel system driven by optical signals addressing the needs of such a physically impaired population is introduced in this paper. The present system is divided into two parts: the first part comprises of detection of eyeball movements together with the processing of the optical signal, and the second part encompasses the mechanical assembly module, i.e., control of the wheelchair through motor driving circuitry. A web camera is used to capture real-time images. The processor used is Raspberry-Pi with Linux operating system. In order to make the system more congenial and reliable, the voice-controlled mode is incorporated in the wheelchair. To appraise the system's performance, a basic wheelchair skill test (WST) is carried out. Basic skills like movement on plain and rough surfaces in forward, reverse direction and turning capability were analyzed for easier comparison with other existing wheelchair setups on the bases of controlling mechanisms, compatibility, design models, and usability in diverse conditions. System successfully operates with average response time of 3 s for eye and 3.4 s for voice control mode.

Low-dose spectral CT reconstruction using image gradient ℓ 0-norm and tensor dictionary.

Spectral computed tomography (CT) has a great superiority in lesion detection, tissue characterization and material decomposition. To further extend its potential clinical applications, in this work, we propose an improved tensor dictionary learning method for low-dose spectral CT reconstruction with a constraint of image gradient ℓ 0-norm, which is named as ℓ 0TDL. The ℓ 0TDL method inherits the advantages of tensor dictionary learning (TDL) by employing the similarity of spectral CT images. On the other hand, by introducing the ℓ 0-norm constraint in gradient image domain, the proposed method emphasizes the spatial sparsity to overcome the weakness of TDL on preserving edge information. The split-bregman method is employed to solve the proposed method. Both numerical simulations and real mouse studies are perform to evaluate the proposed method. The results show that the proposed ℓ 0TDL method outperforms other competing methods, such as total variation (TV) minimization, TV with low rank (TV+LR), and TDL methods.

Filtering Based Adaptive Visual Odometry Sensor Framework Robust to Blurred Images.

Visual odometry (VO) estimation from blurred image is a challenging problem in practical robot applications, and the blurred images will severely reduce the estimation accuracy of the VO. In this paper, we address the problem of visual odometry estimation from blurred images, and present an adaptive visual odometry estimation framework robust to blurred images. Our approach employs an objective measure of images, named small image gradient distribution (SIGD), to evaluate the blurring degree of the image, then an adaptive blurred image classification algorithm is proposed to recognize the blurred images, finally we propose an anti-blurred key-frame selection algorithm to enable the VO robust to blurred images. We also carried out varied comparable experiments to evaluate the performance of the VO algorithms with our anti-blur framework under varied blurred images, and the experimental results show that our approach can achieve superior performance comparing to the state-of-the-art methods under the condition with blurred images while not increasing too much computation cost to the original VO algorithms.

An improved BM3D algorithm based on anisotropic diffusion equation.

Traditional 3D block matching (BM3D) algorithms are among the best denoising methods at present; however, they exhibit the issue of ringing around image edges, which makes them unable to protect image edges and details. Therefore, this paper proposes an BM3D noise processing algorithm for the diffusion equation to reduce image noise without affecting image details, specifically at the edges. This method first uses anisotropic diffusion (AD) filtering for image preprocessing, and then uses the edge direction instead of horizontal direction to search for similar blocks. The AD model is mainly improved to achieve better edges and detailed processing effects. Firstly, with the improved AD direction, a 5 × 5 edge enhancement operator model is implemented in eight directions, and the corresponding gradient information is obtained. This operator improves the processed image edges to achieve clear contours and good continuity. Next, a new calculation method for the diffusion function, whose coefficient is constructed using a hyperbolic tangent function, is introduced. The proposed method is based on the link between the image gradient and diffusion function, and it is mathematically proven that the diffusion function converges faster than the diffusion function of the model proposed by Perona and Malik. Experimental results indicate that the improved model can effectively retain the image edges and texture details, avoid edge ringing, and provide significant improvements in terms of the subjective visual effects and objective numerical indicators.

High-precision image-drift-correction method for EM images with a low signal-to-noise ratio.

The phase correlation method (PCM) is well known for high-precision matching between images. However, if the signal-to-noise ratio of an image is low, the method is difficult to apply. To solve this problem, we developed an improved PCM that can match images automatically with sub-pixel matching precision. Using this method, a 0.2-nm crystal lattice spacing was clearly revealed after 10 blurred images were processed in a verification experiment; such a lattice could not be recognized or hardly be recognized in each individual image.