FG Technische Informatik
Refine
Year of publication
Document Type
Way of publication
- Open Access (6)
Keywords
- ICs (37)
- Zuverlässigkeit (37)
- Selbstreparatur (27)
- Test (24)
- SoCs (22)
- Prozessoren (9)
- Fehlertoleranz (8)
- Encoding (6)
- Fehlersimulation (6)
- Bit rate (5)
Institute
The past years have shown that due to the global success of video communication technology, the corresponding hardware systems nowadays contribute significantly to pollution and resource consumption on a global scale, accounting for 1% of global green house gas emissions in 2018. This aspect of sustainability has thus reached increasing attention in academia and industry. In this paper, we present different aspects of sustainability including resource consumption and greenhouse gas emissions, while putting a major focus on the energy consumption during the use of video systems. Finally, we provide an overview on recent research in the domain of green video communications showing promising results and highlighting areas where more research should be performed.
This paper presents FORTUNE, a hardware-agnostic fault tolerance technique for DNNs that leverages quantization to enhance reliability without significant performance overhead. Unlike conventional methods like Triple Modular Redundancy (TMR), which are computationally expensive, the proposed approach uses memory savings from quantization to protect the critical Most Significant Bit, improving fault tolerance in Deep Neural Networks (DNNs). Memory utilization has been reduced by 37.5% across all networks, with vulnerability in AlexNet reduced by 56% compared to the 8-bit version and 84% compared to the unprotected 3-bit version. These improvements come with only a minor increase in execution time of less than 3%. Using AlexNet as an example demonstrates how our approach effectively enhances memory utilization and resilience while causing only a minimal increase in execution time.
Reliability-aware performance optimization of DNN HW accelerators through heterogeneous quantization
(2025)
The paper introduces a framework for implementing heterogeneous quantization tailored to individual layers of DNN accelerators. The proposed method, evaluated across DNN benchmarks AlexNet, VGG-11 and ResNet-18, enhances both accuracy drop and performance. On average, the accuracy drop at a BER of 10−4 was reduced by nearly 4 times compared to homogeneous quantization, demonstrating a considerable improvement in reliability under fault-prone conditions. Furthermore, the memory overhead introduced by the method for protected heterogeneous networks is less than 0.1%, highlighting that these reliability gains come with negligible resource costs.
Given the growing environmental concerns and significant resource consumption associated with video streaming on electronic devices, measuring the energy consumption is important to guide optimisation and to assess its relative environmental impact. In this paper, we provide comprehensive guidance to accurately measure the energy and power consumption in video communication technologies. We address the complexities inherent in measuring energy consumption across diverse software and hardware setups, with a focus on video communication tasks. We review current measurement techniques, identify limitations in existing practices, and propose a structured methodology that incorporates considerations for static and dynamic power consumption, appropriate sampling frequencies, and statistical rigor. Additionally, we introduce a reference workflow that is adaptable to various multimedia applications and demonstrate its applicability through a case study. By offering clear guidance and practical tools, this work aims to improve the reliability, reproducibility, and comparability of energy consumption measurements in video technologies, providing a strong foundation for the multimedia community to base decisions on.
High-dynamic range (HDR) video content has gained popularity due to its enhanced color depth and luminance range, but it also presents new challenges in terms of compression efficiency and energy consumption. In this paper, we present an in-depth study of the compression performance and energy efficiency of HDR video encoding using High-Efficiency Video Coding (HEVC). In addition to using a native 10-bit HDR encoding configuration as a reference, we explore whether applying tone mapping to an 8-bit representation before encoding can result in additional bitrate and energy savings without compromising visual quality. The main contributions of this work are as follows: 1) a detailed evaluation of four HDR video encoding configurations, three of which leverage tone mapping techniques, 2) a comprehensive experimental setup involving over 15,000 individual encodings across three open-source HEVC encoders (Kvazaar, x265, and SVT-HEVC) and multiple presets, 3) the use of two advanced perception-based metrics for BD-rate calculations, one of which is specifically tailored to capture colour distortions and 4) an open-source dataset consisting of all experimental results for further research. Among the three tone-mapping configurations tested, our findings show that a simple bit-shifting approach can achieves significant reductions in both bitrate and energy consumption compared to the native 10-bit HDR encoding configuration. This research aims to lay an initial foundation for understanding the balance between coding efficiency and energy consumption in HDR video encoding, offering valuable insights to guide future advancements in the field.
With the inception of Industry 4.0, incorporating technologies like the Internet of Things (IoT) into healthcare has become essential. This integration is commonly referred to as the Internet of Medical Things (IoMT). The IoMT is the connection of medical devices using wired or wireless data transmission technology to allow data exchange with the goal of improving the overall healthcare delivery. Despite the numerous advantages that IoMT brings into the healthcare process, there are potential performance challenges that may occur if factors such as data quality and reliability of the IoT devices in different environmental settings are not properly considered. The purpose of this paper is to analyse the performance of connected medical IoT devices that are used for heartrate monitoring based on the aforementioned factors. The setup of the IoMT consists of sensor nodes, which transmit the Electrocardiogram (ECG) data through a multi-protocol gateway to a central server for further data processing. This paper presents the performance analysis of the comparison of four communication technologies: Serial (UART), Bluetooth Low Energy (BLE), Wi-Fi, and 5G NR for real-time ECG monitoring applications, while taking notice of environmental factors that may affect performance. The sensor data transmission is evaluated based on round trip time (RTT) latency, ensuring a desirable throughput and minimal or no data loss. The data readings were taken at varying distances (0.1m to 17m) and sampling rates (300Hz and 1000Hz). The experimental results show that while Serial communication achieves the lowest latency (3.96ms - 4.37ms), Wi-Fi demonstrates consistent Gateway-Server performance (40ms - 60msRTT), 5G excels in short-range communication (1.8ms - 2.0ms Sensor Node-Gateway RTT), and BLE providesbalanced performance (4.86ms - 7.57ms latency). Wi-Fi performed better in long-range scenarios (43.48ms -66.23ms RTT) and maintaining stable performance at longer ranges while 5G shows superior performance in short-range, high-frequency scenarios.
In dynamic industrial environments, strategic sensor placement is key to accurately monitoring equipment and detecting critical events. Despite progress in Industry 4.0 and the Internet of Things, research on optimal sensor placement remains limited. This study addresses this gap by analyzing how sensor placement impacts event detection, using chemical detection as a case study with an open dataset. Detecting gases is challenging due to their dispersion. Effective algorithms and well-planned sensor locations are required for reliable results. Using deep convolutional neural networks (DCNNs) and decision tree (DT) methods, we implemented and tested detection models on a public dataset of chemical substances collected at five locations. In addition, we also implemented a multi-objective optimization approach based on the non-dominated sorting genetic algorithm II (NSGA-II) to identify optimal sensor configurations that balance high detection accuracy with cost efficiency in sensor deployment. Using the refined sensor placement, the DCNN model achieved 100% accuracy using only 30% of the available sensors.
Fixed-point simulation is a critical step in the design flow of Digital Signal Processing (DSP) systems and consumes a significant portion of the overall development time. This work proposes a novel hardware-based approach to improve the performance of fixed-point simulation through a widthconfigurable architecture implemented on Field-Programmable Gate Arrays (FPGAs). To our knowledge, this is the first reported hardware-based solution for fixed-point simulation reported in the literature, which provides substantial speedup compared to traditional software-based approaches. The presented approach utilizes High-Level Synthesis (HLS) tools to create modular, reusable components that enhance the design reusability. This approach accelerates fixed-point simulation for various applications, requiring only minor modifications to the application code to utilize it. By incorporating simulation capabilities within reusable functions, this work improves the efficiency and adaptability of the fixed-point simulation process. Implementation results for two FIR filter applications show that the proposed methodology is superior to traditional software-based approaches in reducing simulation time by two orders of magnitude and thus has the potential to improve the productivity of the DSP system design flow.
Hydrogen (H2) is crucial for replacing fossil fuels and achieving net-zero emissions, but its flammability and explosiveness pose safety challenges. Rapid H2 leak detection is essential for triggering emergency accidents. However, H2 sensor response is constrained by material properties and gas flow dynamics, causing response and detection delays. Our current study explores various available algorithms for H2 sensor response prediction from early responses with a small time window, accelerating leakage detection. Our findings identify the most efficient algorithms for real-time implementation, enhancing H2 safety systems.
Image and signal processing applications have been widely implemented in Field Programmable Gate Arrays (FPGAs) and Graphical Processing Units (GPUs) due to their energy efficiency and performance, respectively. GPUs provide high data processing parallelism and are usually chosen to accelerate applications where low energy consumption is not a high priority. On the other hand, FPGAs are more tailored to hardware solutions due to their reconfigurability, but they struggle to outperform GPUs in data throughput. Soft IP cores implemented on reconfigurable hardware, are an alternative offering advantages from both worlds.
Some of these soft-core solutions offer an entire environment that includes scripts to automate their implementation, custom compilers, and other diverse tools. Unfortunately, some of these soft-cores are dependent on proprietary Intellectual Property (IP) or require hardware expertise to use properly. In this work, we propose an extended version of a popular open-source soft GPU, which can now run alongside a soft RISC-V core, and with High-Bandwidth memory (HBM2) compatibility. Previously, this soft GPU was only ready to be deployed in boards with a hard ARM core, but now it can be easily used in FPGAs without this requirement. We also provide an evaluation of how the soft GPU performs with respect to the pure RISC-V core, and a hard ARM core achieving geometric mean speed-ups of 114.60x and 19.72x respectively when performing some image and signal processing applications. Finally, we demonstrate how our soft GPU benefits from the HBM integration.