Albrecht Rothermel

Clock/data recovery PLL using half-frequency clock

A CMOS clock and data recovery PLL is described for serial nonreturn-to-zero (NRZ) data transmission. The voltage controlled oscillator (VCO) works at half the data rate, which means for a 1-Gb/s data rate, the VCO runs at 500 MHz. A specially designed phase comparator uses a delay-locked loop (DLL) to generate the required sampling clocks to compare clock and data. The VCO can typically be tuned from 350 MHz to 890 MHz, and the phase-locked loop (PLL) locks between 720 Mb/s and 1.3 Gb/s. Data recovery is error free up to 1.2 Gb/s with a 9-b pseudorandom data sequence. The core consumes 85 mW (3.3 V) at 1 Gb/s.

A CMOS Chip With Active Pixel Array and Specific Test Features for Subretinal Implantation

This paper presents a CMOS imager chip that is aimed at subretinal implantation for partially restoring human vision. It has low supply voltage (plusmn 2 V) and all DC free terminals for long life wired operation. Stimulation voltage is increased to approximately 4 Vpp by low voltage drop design. 40 x 40 pixel cells including light sensors, amplifiers, control logic and electrode drivers are addressed sequentially to improve power consumption and spatial resolution of perception. Pad count is limited to 6, which requires a specific test procedure. The 3 x 3 mm2 design is fabricated in a 0.35 mum CMOS technology optimized for optical performance.

Motion adaptive intra refresh for the H.264 video coding standard

An intra refresh algorithm is presented which enables low-delay video communication at the efficiency of standard IP-coding schemes. It guarantees a complete recovery of erroneous transmission within a defined time interval at low computational complexity.

Motion adaptive intra refresh for low-delay video coding

An intra refresh algorithm is presented which enables low-delay video communication at the efficiency of standard IP-coding schemes. It guarantees a complete recovery of erroneous transmission within a defined time interval at low computational complexity.

Learning convolutional neural networks from few samples

Learning Convolutional Neural Networks (CNN) is commonly carried out by plain supervised gradient descent. With sufficient training data, this leads to very competitive results for visual recognition tasks when starting from a random initialization. When the amount of labeled data is limited, CNNs reveal their strong dependence on large amounts of training data. However, recent results have shown that a well chosen optimization starting point can be beneficial for convergence to a good generalizing minimum. This starting point was mostly found using unsupervised feature learning techniques such as sparse coding or transfer learning from related recognition tasks. In this work, we compare these two approaches against a simple patch based initialization scheme and a random initialization of the weights. We show that pre-training helps to train CNNs from few samples and that the correct choice of the initialization scheme can push the networks performance by up to 41% compared to random initialization.

Automatic Databases for Unsupervised Face Recognition

In this paper, an automatic system is presented to establish databases for face recognition from video. We propose a temporal-based face detector, which can improve the detection rate as well as the recognition rate when working together with an image-based face detection method. Several adaptive thresholds and a filter are introduced to improve face recognition performance and to keep the purity, variety and rapidity of the face databases. The system deals with multiple coexisting persons without requirement for any user interaction, and without requirement for any human supervision or assistance.

Interim Results of a Multicenter Trial with the New Electronic Subretinal Implant Alpha AMS in 15 Patients Blind from Inherited Retinal Degenerations

Purpose: We assessed the safety and efficacy of a technically advanced subretinal electronic implant, RETINA IMPLANT Alpha AMS, in end stage retinal degeneration in an interim analysis of two ongoing prospective clinical trials. The purpose of this article is to describe the interim functional results (efficacy). Methods: The subretinal visual prosthesis RETINA IMPLANT Alpha AMS (Retina Implant AG, Reutlingen, Germany) was implanted in 15 blind patients with hereditary retinal degenerations at four study sites with a follow-up period of 12 months (www.clinicaltrials.gov NCT01024803 and NCT02720640). Functional outcome measures included (1) screen-based standardized 2- or 4-alternative forced-choice (AFC) tests of light perception, light localization, grating detection (basic grating acuity (BaGA) test), and Landolt C-rings; (2) gray level discrimination; (3) performance during activities of daily living (ADL-table tasks). Results: Implant-mediated light perception was observed in 13/15 patients. During the observation period implant mediated localization of visual targets was possible in 13/15 patients. Correct grating detection was achieved for spatial frequencies of 0.1 cpd (cycles per degree) in 4/15; 0.33 cpd in 3/15; 0.66 cpd in 2/15; 1.0 cpd in 2/15 and 3.3 cpd in 1/15 patients. In two patients visual acuity (VA) assessed with Landolt C- rings was 20/546 and 20/1111. Of 6 possible gray levels on average 4.6 ± 0.8 (mean ± SD, n = 10) were discerned. Improvements (power ON vs. OFF) of ADL table tasks were measured in 13/15 patients. Overall, results were stable during the observation period. Serious adverse events (SAEs) were reported in 4 patients: 2 movements of the implant, readjusted in a second surgery; 4 conjunctival erosion/dehiscence, successfully treated; 1 pain event around the coil, successfully treated; 1 partial reduction of silicone oil tamponade leading to distorted vision (silicon oil successfully refilled). The majority of adverse events (AEs) were transient and mostly of mild to moderate intensity. Conclusions: Psychophysical and subjective data show that RETINA IMPLANT Alpha AMS is reliable, well tolerated and can restore limited visual functions in blind patients with degenerations of the outer retina. Compared with the previous implant Alpha IMS, longevity of the new implant Alpha AMS has been considerably improved. Alpha AMS has meanwhile been certified as a commercially available medical device, reimbursed in Germany by the public health system.

Evaluation of compression algorithms for automotive stereo matching

Automotive stereo-based camera systems are at the focus of new vehicle functions development. Advantageous interconnections between automotive cameras and electronic control units are possible using IP-based technology in the vehicles. Because automotive implementation of this technology allows only limited data rates camera stream compression is necessary to transmit over such a medium. The image processing algorithms can be applied after a decompressing of the camera streams. This paper introduces a method for comparing the impact of different image compression algorithms on automotive stereo algorithms. We propose a comparison metric based on disparity error to compare the effects as a function of image compression algorithms, compression ratio and vehicle velocity.

In vitro study of titanium nitride electrodes for neural stimulation

For neural stimulation, reliable high density charge transfer into tissue is required. One electrode material for these applications is titanium nitride (TiN). In this paper, a method for lifetime analysis of TiN electrodes is discussed. Our method significantly differs from open literature. The tests were run for much longer durations. Special attention was paid to the optical appearance and electrode voltage response to different input current pulses. According to our investigations, TiN electrodes are able to deliver at most 0.2mC/cm2 charge density for square shaped electrodes with 50μm × 50μm dimensions in safe operation, which is less compared to previous reports. The safe operation window for TiN was confirmed to be ±1V in terms of electrode potential with the counter electrode considered as reference. We found that the shape of the waveform does not affect electrode lifetime. Our measurements show that rectangular voltage waveforms inject the most amount of charge into the electrodes compared to other shapes. This makes rectangular electrode voltage signals optimal for highest charge injection at a given lifetime. In our case with square electrodes, the absolute electrode potential is found to be the more important parameter in electrode lifetime, compared to Helmholtz capacitor voltage drop.

Reduction of computational cost for high quality video scaling

Due to the increasing number of video and image resolutions, scaling gets more and more important for image quality. When it comes to mobile entertainment, e.g. in cars or for cell phones, a number of restrictions apply. For example, space, weight, and power consumption are limited which requires limitations on processing power and memory. In this paper, we introduce an adaptive hybrid decision algorithm for video or image scaling with method-percentage control (MPC). The algorithm combines a simple interpolation method with a more computational cost intensive one. With this approach, it is possible to maintain the quality of a high level interpolation while saving computational costs of up to 66% with respect to a high quality scaling algorithm. With these characteristics it is well qualified for mobile applications.