Eye tracking technique is showing broad application potential in daily life. This work proposes an FPGA-based eye gaze tracking system with a differential convolutional neural network structure, which is naturally compatible with calibration operation. Each newly acquired eye image is fed into the proposed algorithm together with the pre-calibrated eye sample as a reference to obtain the relative gaze direction. The final processing output is generated by combining the results of several reference samples. Our design reports an angular gaze tracking accuracy of 3.28° and 3.05° on MPIIGaze and a self-built dataset, respectively. The entire system is integrated on a Xilinx Ultra 96 single computer board with FPGA for acceleration. The total weight of the board is 123.3 grams. The system achieves a processing speed of 15.2 frames per second.

This paper introduces a tangible user interface based on projector-camera system. The interface is encoded by stripes in an imperceptible form. We make the first attempt to employ a neural network for extracting hand information from structured light images. The lightweight network of only 0.24 MB parameters infers hand existence, hand segmentation and fingertip locations. These results are utilized for decoding the imperceptible stripes. Experimental results show that the proposed network outperforms state-of-the-art methods, achieving a high IOU of hand segmentation (0.996), and fingertip localization precision (98.98%) within a 10-pixel range. The interactive system performs a touch detection precision of 98.78%, with an average fingertip localization error of 3.33 pixels. It supports real-time interaction with an average execution time is 24.87ms.

Camera calibration is a crucial step in many computer vision applications that require 3D information about the scene. The process involves estimating intrinsic and extrinsic parameters from a set of captured images. The main challenge is to obtain accurate corner position information from the captured patterns. The conventional method uses only one checkerboard pattern, which suffers from uneven noise distribution. To solve this problem, a calibration method based on complementary patterns is proposed, which reduces the influence of image noise and enhances the calibration accuracy. Experimental results show that the proposed method achieves a 27% reduction in the Root Mean Square of the re-projection error compared with the non-complementary pattern.

In shallow underwater environments, active sonar is often utilized for the detection of small moving targets. However, such systems can suffer from high-level background reverberation, making it challenging to detect and distinguish the target echoes from reverberation. To address this challenge, we propose a fast moving object detection method utilizing multi-scale spatial-temporal lacunarity. Specifically, computing lacunarity solely from temporal or spatial dimensions makes it difficult to distinguish between target and reverberation patterns. Our method overcomes this by characterizing both the static and dynamic patterns of target echoes and background reverberation through spatial-temporal lacunarity. Additionally, we introduce an echo pyramid that enables multi-scale observation while reducing computational complexity. Experimental results demonstrate that our proposed method significantly improves the detection of small moving targets from high-level background reverberation. Furthermore, our method outperforms existing methods in terms of visual quality and quantitative metrics.

This paper proposed a wireless 16-channel im-plantable system for long-term neural recording. In order to achieve stable and reliable neural signal acquisition, the im-plantable microsystem consists of an analog front end (AFE), a transmitter (TX), a small battery and a coil for energy harvesting to charge the battery. The AFE integrates 16-channel low-noise amplifiers (LNA), a SAR ADC and a digital interface. The TX integrates a rectifier, a bandgap voltage reference, regulators and a 427 MHz OOK modulated transmitter. The AFE and TX chips were fabricated in 180-nm technology. All the required functional modules are integrated in the chips with off-chip crystal, coil and antenna. The proposed microsystem weighs 2.1 g without a battery, and 3.7 g including the battery. The dimension is$20\times 18\times 7$mm 3. The total current of the system is 1.13 mA, and the battery life is about 50 hours with a capacity of 60 mA$h$. The charging current is 10mA under wireless power transmission.

This paper proposed a fully integrated MBAN (2360–2400 MHz) continuous phase modulated transmitter (TX) with tunable less than 0dBm output power for medical band. A duty-cycle tuning strategy was proposed for the power amplifier (PA) featuring adaptive optimized efficiency for different output powers. A fully on-chip transformer-based match network was proposed to suppress the 2nd harmonic using a series$LC$resonator and to suppress the 3rd harmonic by introducing a transformer inter-winding capacitor feedback path. A fractional-N all-digital phase locked loop (ADPLL) with a transformer-based digitally controlled oscillator (DCO) is employed to reduce power consumption as well as improve modulation quality. The transmitter was fabricated in 40-nm CMOS technology, occupying an active area of 0.48mm2. Experimental results show a 26% drain efficiency with −10dBm PA output and 4dB tunable range. A 2mW total power consumption was measured with a TX efficiency of 5% and an energy efficiency of 2nJ/bit. The measured 2nd and 3rd harmonic distortion of the output were −44.3dBm and −57.2dBm, respectively, with on-chip matching network. The measured FSK error of CPM was 2.3% with an M of 2 and 1.57% with an M of 4.

Stripe extraction is a deciding factor in the multi-line structured light systems. This paper introduces a method to extract the deteriorated stripes in a robust and accurate way. Based on the photoelectron crosstalk phenomenon of image sensors, mathematical expression of the ideal cross-section distribution for the captured stripes is deduced. Then, we establish a probability model for each pixel, with the consideration of how well the local distribution fits the ideal distribution, and how possible it is linked with the previous stripe centers. Therefore, the extraction process is transformed into a local searching scheme with dynamic update. Experimental results show that the robustness of the proposed scheme outperforms other methods. The root mean squared error of the reconstructed point cloud is 0.219mm, which proves the validity of the proposed method.

Forward-looking sonar (FLS) often suffers from complex underwater environments. It is hard to detect small objects from the FLS imagery characterized by low signal-to- noise ratio and low resolution. To highlight the object from severe noise background, we propose an object enhancement method for objects in the regions of interest (ROIs) based on multi-frame image fusion. This method includes two crucial steps: 1) A Fourier-based multi-stage registration algorithm is proposed to solve the problem of a drastic change of object position between frames due to long target distance and rapid change of azimuth angle. 2) A multi-frame fusion algorithm based on self-supervised deep learning is adopted to enhance the ROIs. Experimental results demonstrate that our proposed enhancement method can significantly highlight the objects in the ROIs and has excellent noise suppression in terms of quantitative metrics and visual quality.

The discrete-time amplifier has the property of sample and hold, thus it is difficult to adopt local feedback method to deal with offset and flicker noise. A calibration-free swap-combine noise cancellation technique is proposed to remove the input-pair-introduced mismatch in discrete-time amplifiers with keeping the same external interface so it can directly replace the original amplifier. The proposed method can eliminate the flicker noise by modifying the noise transfer function, as well as achieve significant offset suppression. Simulation of a prototype shows that it can reduce the offset root mean square from 10mV to 0.25mV. Benefiting from avoiding calibration loops, the proposed method can also deal with input common-mode voltage variation and temperature drift, as well as reducing the calibration area cost and initial delay.

An ultra-wide band, high linearity and low phase noise 43 GHz voltage-controlled oscillator (VCO) for frequency-modulated continuous-wave (FMCW) radar system is presented. A transformer-based magnetic-tuning technique is adopted to realize both coarse frequency tuning and fine frequency tuning, achieving a sweeping bandwidth of >5.7 GHz within single tuning curve. Distributed biasing parallel varactors are used to achieve high tuning linearity. Since the flicker noise (1/f noise) significantly worsens the VCO phase noise due to high 1/f corner frequency in TSMC 28nm CMOS, a technique to suppress the flicker noise up-conversion is adopted to improve the phase noise performance. The VCO has been implemented in TSMC 28 nm CMOS process, obtaining a tuning range (TR) from 37.5 GHz to 49 GHz, which consumes 8 mA current from a 0.95 V power supply. The phase noise is lower than -100 dBc/Hz at a 1 MHz offset from a 37.5 GHz carrier and the resulting FoM is -182 dBc/Hz.

This paper proposes a static hand gesture recognition method with low computation and memory consumption for wearable applications. The hand contour is chosen as the hand gesture feature and support vector machine is used to classify the feature. A multiscale weighted histogram of contour direction-based direction normalization is proposed to ensure good recognition performance. In order to improve efficiency, the proposed histogram only counts the direction of the contour point to focus on the most significant hand feature in the first-person view of wearable devices. Based on the hand's anatomy, the proposed histogram is weighted by considering each contour point's position and direction jointly using the direction-angle map, to ensure robustness. Experimental results show that the proposed method can give a recognition accuracy of 97.1% with a frame rate of 30 fps on a PC.

A hand segmentation framework is proposed for 3-D hand gesture interaction for wearable devices. In this framework, all the objects in a scene are regarded as directed trees in a forest, and the problem of the hand segmentation can be converted into finding the target tree (called hand tree) in the forest with proper hand properties including color consistency, space consistency, disparity, and hand shape constraints. The forest grows scan-line by scan-line from high reliable regions to low reliable regions. First, the vertices in each tree are generated through boundary detection. To get rid of the interference of skin-colored objects, an under-segment vertex splitting step is added in. Second, a scan-line-based forest growing method is proposed by constructing the maximum spanning forest (MSF) with the edges weighted by the consistencies of color and space. Meanwhile, the high-order history information of the tree construction propagates along each growing tree to the current scan-line to predict the 3-D location of the growing hand tree, so as to achieve semiglobal optimization. At last, the growing hand tree is trimmed and labeled with the disparity and hand shape constraints to correct the misconnection in the MSF. The vertex's disparity is obtained by our multipriority vertex stereo matching algorithm using the 3-D location prediction to overcome the problem of occlusion and high similarity between fingers. Experimental results demonstrate that by using our method, the hand can be well segmented. The score and accuracy of the hand segmentation result is about 96.3% and 92.9%, respectively.

Robot-assisted needle insertion technology is now widely used in minimally invasive surgeries. The success of these surgeries highly depends on the accuracy of the position and orientation estimation of the needle tip. In this paper, we proposed an optical tracker based registration method for the marker-robot-needle tip registration in a robot-assisted needle insertion surgery. The method consists of two steps. First, with the guide of the optical tracker, we use the motion vector from the marker's current position to its target point as a feedback control to automatically register the robot-marker and the robot-tracker coordinates respectively. Then, in the procedure of marker-needle tip registration, we use two points on the needle holder to achieve a high accuracy needle orientation registration. We have conducted a series of experiments for the proposed method on a verified platform. Results show that the position estimation error is below 0.5mm and the orientation estimation error is below 0.5 degrees.

This paper proposes a method to build a bare-finger Projector-Camera-Touchpad (PCT) using only one projector and one camera to make any plane become a touchpad. A finger model composed of an elliptic cylinder representing finger pulp, and a quarter ball as the fingertip, is introduced to help fit the finger's pose and size for accurate touch detection and motion tracking. We adopted the color structured light composed of just two kinds of colors, i.e., cyan and magenta, to get depth information of the points on projection pattern for the PCT application. Experiment results show that the accuracy of the touch detection is almost 99% when the finger is under 3mm away from the projection plane and the motion tracking accuracy is under 1 pixel per centimeter, which can satisfy most customers demand. The PCT works just like a real touchpad.

In this paper, we propose an accurate automatic chessboard corner detecting algorithm of sub-pixel precision. The algorithm is based on a novel hyperbolic tangent model. A set of corners are detected initially by Harris' algorithm. Then the hyperbolic tangent model for each corner point is optimized by parameter adjustment. Finally, we use the optimal model to obtain the sub-pixel position of each corner with fake corner elimination. Experiments on three datasets show that the proposed method achieves a detection rate of at least 98% and a reprojection error in camera calibration of at most 0.85 pixels. Our algorithm significantly outperforms the methods in three related works.

This paper presents a method to design a high efficiency energy receiving coil for Micro-Ball endoscopy. The remained space for energy source is extremely small (below 6.6∗6.6∗6.6mm3) in Micro-Ball. Ferrite core is adopted to improve the coupling coefficient. Based on detailed analysis of power loss in the receiving coil, parameters such as number of turns, frequency, types of wires (solid wire or litz wire) was optimized to improve the quality factor of the receiving coil. Ultimately, the coupling coefficient has an improvement of 80%. Experiments shows that the quality factor of the receiving coil (a=b=6.6mm, h=6.6mm) can reach 47 with solid wire and 96 with litz wire at 1.5MHz. Hybrid resonant is proposed to fit in with the load of the Micro-Ball. The AC-AC transmission efficiency of the receiving coil with litz wire is 6.3% in the worst case.

This paper presents a combined wireless power transmitting coil designed for endoscopic capsule. In the design, a group of vertically segmented transmitting coils are combined to work. According to the position of the capsule, one, two or three segmented transmitting coils will be selected to work, which makes the capsule closer to the center of the combined transmitting coils. The mechanism ensures high power transfer efficiency and good homogeneity of the energy distribution. This paper also gives the theoretical expression about how to combine the segmented transmitting coils to work together with the highest performances. The experiments show that the minimum and average AC-AC transfer efficiency are 7.53% and 12.7% respectively.

This paper proposes a new fingertip touch detecting method with high accuracy for the human-computer interaction based on one projector and a single camera. A finger model composed of a quarter-sphere and half a cylinder is introduced, and the relationship between the finger contour and its shadow based on this model is given out. The way to estimate the distance between finger and the projected surface is presented and discussed. Experimental results show that our approach can achieve a precision of about 2mm when the distance from the projected surface to the projector is 50cm and the button size will not influence the precision.

In this paper, an intestinal lumen detection method suitable for hardware design is proposed for the locomotive capsule or intelligent Micro-Ball endoscopy. In order to get low complexity as well as relatively high detection precision, a global threshold segmentation algorithm for non-full-shrink lumen and texture based algorithm for full-shrink lumen are proposed. Furthermore, only the red channel of the original Bayer-format image is used in the two algorithms. The computational complexity of the lumen detection method is also given. Experimental results show this method can provide 99.3% precision and 99.0% sensitivity.

This letter proposes an efficient two-step segmentation method for large-scale 3-D point cloud data collected by the mobile laser scanners. First, a new scan-line-based ground segmentation algorithm is designed to filter the points corresponding to the ground with high accuracy. Second, we propose a selfadaptive Euclidean clustering algorithm to further separate the off-ground points corresponding to different objects. Experiments show that our method delivers superior segmentation results on scanned data. In fact, the proposed method can be used in complex scenes including slope and bumpy road at an error rate of 0.674% and a computing throughput of over 20 million points/s.

In this paper, we propose a novel interactive projection system (IPS), which enables bare-finger touch interaction on regular planar surfaces (e.g., walls, tables), with only one standard camera and one projector. The challenge of bare-finger touch detection is recovering the touching information just from the 2-D image captured by the camera. In our method, the graphical user interface (GUI) button is projected on the surface and is distorted by the finger when clicking it, and there is a significant positive correlation between the button's distortion and the finger's height to the surface. Therefore, we propose a novel, fast, and robust algorithm, which takes advantage of the button's distortion to detect the touch action. The proposed touch detection algorithm is performed in three stages: 1) region of interest extraction through a homography mapping, by which the computational complexity of the following processing is reduced; 2) the button's distortion detection using a special edge detection algorithm, which greatly reduces the errors due to the influence of the finger's shadows and edges; and 3) touch action judgment by the button's distortion. Several applications (e.g., virtual keyboard, power point viewing), which use the proposed touch detection method based on the buttons, are shown in this paper. An evaluation is performed on the virtual keyboard and the results demonstrate that the proposed approach can detect bare-finger touch in real time with the missed detection rate of 1.00%, false detection rate of 2.08%, and touch detection rate of 96.92% at the typical projected distance.

This paper proposes a new wireless power transfer (WPT) system for the endoscopic capsule. In the system, a group of vertically segmented primary coils around the human body are adopted as the transmitting coils. According to the position of the endoscopic capsule, one of the transmitting coils is intelligently selected to transfer the power with the highest efficiency. This paper also presents a coil selection and control algorithm for the unique WPT system. Compared with the conventional design, the calculated transmission efficiencies of the new solution at the best and worst points of the body are enhanced by 53.5% and 138% respectively. And the average efficiency of the proposed solution is about 3.8%.

This paper firstly proposes a new algorithm for finger touch detection based on front-projected vision-based table system. The algorithm, taking advantage of shadow information, can locate possible finger touch with low complexity. An adaptive structured light method is introduced to improve the precision of touch detection effectively. Secondly, we present a new prototype system architecture and corresponding hardware design for portable bare-hand human computer interaction (HCI). Experimental results show that the accuracy of touch detection of single finger and multi-finger is 98.3% and 92.6% respectively.

A chipset, including one master chip and six same slave chips, is designed and implemented for the endoscopic Micro-Ball, which has six cameras to minimize the blind area while examining the human gastrointestinal (GI) tract. For lowering power consumption, a multi-level clock management for the whole chipset is implemented. A power and area efficient 4 × 4 JPEG image compressor with high image quality is integrated into the slave chip to reduce energy consumption and save FLASH memory space. The master chip and slave chip have been fabricated in 0.18 μm CMOS technology. The experimental results show that the designed chipset for Micro-Ball can support different image frame rates: from 1fps to 6fps with 480×480 image resolution. The power consumption of the chipset is 1.6mW@2fps. The power consumption of the Micro-Ball is estimated about 6mW@2fps. It can work for more than 15 hours, powered by a 1.55V@60mAh battery.

This paper proposes a theoretic error analysis framework as well as corresponding calibration methods for a mobile 3D data acquisition system equipped with LiDAR and GPS/IMU. Our framework considers inherent errors of variable sensors, installation errors, and time synchronization errors among different sensors. Our framework is enhanced by an empirical error analysis method based on a triangulated irregular network structure. Experimental results prove that the proposed error analysis methodology can provide accurate and efficient evaluation for mobile 3D data acquisition systems.

Attitude sensing system is indispensable for wireless Micro-Ball endoscopy to capture the full-view images of gastrointestinal (GI) tract. This paper proposes a novel attitude sensing system design with small size and low power for this application, which is based on MEMS 3-axis accelerometer and magnetometer. To reduce the error caused by tilt angles, a compensatory algorithm is introduced in this design. The experimental results show that the attitude sensing system could measure attitude in real time with high enough accuracy meeting the requirement of the wireless Micro-Ball endoscopy application. The power consumption of the attitude sensing system is about 3mW. The error of pitch is less than 3°, roll is less than 3.1°, and yaw is less than 6°.

The wireless endoscopic Micro-Ball is used to capture images of gastrointestinal (GI) tracts and save them into Flash memories. The batteries will be exhausted after the Micro-Ball is excreted from human body. So a new wireless power transmission (WPT) solution is proposed to deliver energy from an image reader to the Micro-Ball to fetch the stored images. In this solution, considering that the posture of the Micro-Ball is uncertain when it is placed into the image reader and the volume of the Micro-Ball is too small to contain multiple receiving coils, we employ multiple emitting coils in the image reader and one single receiving coil inside the Micro-Ball. Additionally, an adaptive control mechanism is proposed to select the optimum emitting coil with the highest power efficiency to transmit wireless power to the Micro-Ball placed inside the image reader. As a result, the Micro-Ball can work well in the image reader under all postures with highest power efficiency. The experimental results prove that the power delivering efficiency is in the range of 32% to 36%.

Nowadays the interpretation of the images acquired by wireless endoscopy system is a tedious job for doctors. A viable solution is to construct a map, which is a 2D representation of gastrointestinal (GI) tract, based on image registration to reduce the redundancy of images and improve the understandability of them. Nevertheless, the limited field of view makes it difficult for traditional wireless capsule endoscopy system to capture sufficient image information for a complete GI map. The work reported in this paper addresses the problem of the 2D representation of GI tract based on a new wireless Micro Ball endoscopy system with multiple image sensors. This paper analyses the characteristics of images captured by Micro-Ball endoscopy system and proposes a determination method for registration of wireless endoscopic images based on phase correlation method (PCM). The performance of PCM and the proposed determination method is verified with experiments.

By analyzing the features unfit for parallel computation and low power implementation, a VLSI architecture of JPEG-LS encoder for lossless image compression is proposed in this paper. It functionally consists of four parts: Mode decision module, clock controller, three linear parallel pipelines, and a two-tier data packer. Computations are organized in a fully pipelined style in these modules, so that real time data processing can be achieved. The clock management scheme with four interlaced clock domains and a dedicated clock controller is applied to ensure the bottleneck calculation, reduce the clock frequency on non-critical paths, and shut off the working clocks of idle modules, which reduces 15.7% of overall power consumption. The proposed JPEG-LS encoder with the features of low power and high processing speed, has been applied in a wireless endoscopy system.

A design of low power digital baseband for wireless endoscope capsule is presented. The key design issues involved in this IC are discussed, including implementation of communication protocol, real time image filter and real time JPEG-LS encoder. Power dissipation is lowered through the architectural exploration. The baseband has been implemented in 0.18 μm CMOS technology. Measurement results show that 50% power reduction is achieved when image compression is enabled @ 30 fps for VGA image at 1.2 V supply voltage compared with the situation when image compression is disabled, 11% power reduction is achieved compared with the previous research.

This paper proposes a new low complexity near-lossless image compression method and its VLSI design for low power and high frame rate in the wireless endoscopy capsule system. Assuring outstanding compression performance and high image quality, the proposed method with the features of low complexity, low storage overhead and real time data processing, makes it ideal for hardware implementation. The VLSI architecture consists of two pipelined parts: The preprocessor and the JPEG-LS engine. A fully pipelined VLSI structure with a dedicated clock management scheme is proposed for the JPEG-LS engine, which ensures a low power application, and real time data processing as well. The hardware implementation has been verified on FPGA and implemented in 0.18μm CMOS technology.

This work presents the design and implementation of a 2.4GHz low power wireless transceiver analog front-end for the endoscopy capsule system in 0.25/spl mu/m CMOS process. The prototype integrates a low-IF receiver analog front-end (low noise amplifier, double balanced down-converter, band-pass filtered AGC loop and ASK detector) and a direct up-conversion transmitter analog front-end (20MHz IF PLL with well defined amplitude control circuit, ASK modulator, up-converter and output buffer) on a single chip together with an internal RF local oscillator and LO buffers. Design trade-offs have been made over the boundaries of the different building blocks to optimize the overall system performance. All building blocks feature circuit topologies that enable comfortable operation at low power consumption. As a result, the IC operates from a power supply of 2.5 V, while only consuming 15mW in receiver (RX) mode and 14mW in transmitter (TX) mode.

In order to reduce the communication bandwidth and the transmitting power in the wireless endoscopy capsule, a high quality image compression algorithm has to be used. This paper proposes a new VLSI structure for the improved near-lossless color image compression algorithm to reduce the memory consumption 47% and save the power in the wireless endoscopy capsule. The improved compression algorithm can provide low average compression rate (2.12bits/pixel) with high image quality (larger than 53dB) for endoscopic images. The VLSI structure has been implemented in 0.18 /spl mu/m CMOS process.

This paper presents a new method for near-lossless image compression for digital colorful image sensors with Bayer color filter arrays (CFA). In this method, the captured CFA raw data is first smoothed by low-pass filters followed by down-sampling and then compressed directly before full color interpolation which introduces redundancy. Assuring high image quality, the method can provide higher compression ratio and lower complexity than conventional image compression methods and other existing similar methods. The composite peak signal to noise ratio (CPSNR) of the decompressed image is larger than 51 dB.

In order to decrease the communication bandwidth and save the transmitting power in the wireless endoscopy capsule, this paper presents a new near-lossless image compression algorithm suitable for hardware design based on the Bayer format image. This algorithm can provide low average compression rate (2.12 bits/pixel) with high image quality (larger than 53.11 dB) for endoscopic images. Especially, it has low complexity hardware overload and supports real-time compressing. In addition, the algorithm can provide lossless compression for the region of interest (ROI) and high quality compression for other regions. The ROI can be selected arbitrarily by varying ROI parameters.

This paper presents a new approach for near-lossless and lossless image compression in digital colorful image sensors with Bayer color filter arrays (CFAs). In this approach, the captured CFA raw data is firstly transformed from quincunx shape to rectangular shape and then smoothed by a low-pass filter. Lastly, the filtered data are compressed directly before full color than conventional interpolation-first image compression methods and other existing similar compression-first methods. Furthermore, through altering the quality control factor, the image quality, PSNR, can be changed from 44.15 dB to infinite with the compression ratio from averagely 3.3 bits/pixel to 6.9 bits/pixel.