SBIR Phase I: Ultra Power-efficient Biologically-Inspired Integrated Circuit Architectures for the Processing and Classification of Analog Sensor Signals

SBIR 第一阶段：用于模拟传感器信号处理和分类的超节能仿生集成电路架构

• 批准号: 1346123
• 负责人: Tom Darbonne
• 金额: $ 14.97万
• 依托单位: MAVRIC Semiconductor Inc
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-01-01 至 2015-10-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1346123&HistoricalAwards=false
• 关键词: SBIR; Phase; Ultra; Power; efficient

This Small Business Innovation Research (SBIR) Phase I project demonstrates the disruptive energy-efficiency advantages of neural classifier approaches for context aware applications (embedded systems incorporating a multiplicity of sensors to autonomously infer their current state in noisy conditions). The value proposition benefits a wide range of consumer (wearable computing), telecommunication (mobile handset), industrial (internet-of-things edge nodes), medical (eHealth, portable medical devices, implanted devices), and military (autonomous control drones) applications which incorporate multiple sensors and where battery life/small form factor/product weight/device accuracy trade-offs are concerns. We address the digital signal processing power consumption problem, which requires engineers designing battery-powered systems to make undesirable product trade-offs. Our proprietary mixed signal approaches uniquely close this traditional power/performance gap, solving a long-standing challenge for battery-powered/autonomously powered devices. The broader impact/commercial potential of this project includes disruption of traditional design approaches for battery-powered intelligent devices, accelerated scientific research, more efficient mixed signal electronic curriculum, new classes of portable devices that enhance the human condition and extend healthy life, and job creation. This project will demonstrate fundamental breakthroughs which can subsequently be extended to create a new generation of low-power configurable computing devices with potential impact matching the commercialization of earlier break-through digital technologies (FPGA, GPU, embedded processor). Over the past 30 years, handheld electronics have improved energy efficiency 1000X; our approach enables another factor of 1000 improvement in energy efficiency for potential applications. With advances in the tool-chain, experimental hardware from diverse scientific disciplines (neuroscience) incorporating FPAA technology will implement formerly impractical algorithms, achieving fresh scientific insights. Eventual tool enhancements support opportunities in engineering and computer science laboratory sensor-related curriculum. The initial wearable computing target market was based on strong technology-market fit, accelerated product adoption cycles, and a $300M market opportunity. Wearable devices incorporating FPAA technology will be easier and safer to use, less obtrusive, and communicate more accurately.

这个小企业创新研究（SBIR）第一阶段项目展示了神经分类器方法在上下文感知应用（嵌入式系统包含多个传感器，可以在嘈杂的条件下自主推断其当前状态）中的破坏性能效优势。 该价值主张使广泛的消费者（可穿戴计算）、电信（移动的手机）、工业（物联网边缘节点）、医疗（电子健康、便携式医疗设备、植入设备）和军事（自主控制无人机）应用受益，这些应用包含多个传感器，并且关注电池寿命/小尺寸/产品重量/设备精度权衡。我们解决了数字信号处理功耗问题，这需要设计电池供电系统的工程师做出不受欢迎的产品权衡。 我们专有的混合信号方法以独特的方式缩小了这一传统的功率/性能差距，解决了电池供电/自主供电设备的长期挑战。该项目的更广泛的影响/商业潜力包括打破电池供电智能设备的传统设计方法，加速科学研究，更有效的混合信号电子课程，提高人类条件和延长健康寿命的新型便携式设备，以及创造就业机会。该项目将展示基本突破，随后可以扩展到创建新一代低功耗可配置计算设备，其潜在影响与早期突破性数字技术（FPGA，GPU，嵌入式处理器）的商业化相匹配。 在过去的30年里，手持电子产品的能源效率提高了1000倍;我们的方法使潜在应用的能源效率提高了1000倍。 随着工具链的进步，来自不同科学学科（神经科学）的实验硬件结合FPAA技术将实现以前不切实际的算法，实现新的科学见解。最终的工具增强支持工程和计算机科学实验室传感器相关课程的机会。最初的可穿戴计算目标市场是基于强大的技术市场契合度、加速的产品采用周期和3亿美元的市场机会。 采用FPAA技术的可穿戴设备将更容易使用，更安全，更不显眼，并且更准确地进行通信。