SBIR Phase II: Neural Algorithms for Multimodal Sensory Analysis

该项目更广泛的影响/商业潜力是在广泛的能量受限传感器设备中实现连续传感应用。如果在超低功耗感官处理的发展方面没有显著的创新，连续和精确的传感将仍然是一个小众应用，局限于具有稳定和丰富的电源和大量计算资源的环境。本提案中描述的技术将展示在具有严格能源限制的移动或远程环境中广泛部署连续传感设备的可行性和潜力。智能手机和可穿戴设备市场是该技术的一个重要直接市场，拥有庞大的客户群，许多新兴的终端用户应用程序可以利用环境感知来触发基于情境和预期的行动。拟议的技术广泛适用于许多其他市场和领域，包括关键患者传感器的医疗、健康和安全监测、个人健身设备、军事应用和环境监测器。为这些和其他应用灵活部署连续传感的能力有可能彻底改变这些市场，并创造全新的、不可预见的应用领域。这个小企业创新研究第二阶段项目计划研究和开发基于生物脉冲神经元特性和人类大脑感官处理能力的算法。人类大脑的独特之处在于，它能够使用一个基本的计算元素——脉冲神经元，并执行各种各样的任务。大脑有能力从多种方式（触觉、视觉等）中准确地分类感觉模式，解释外部世界，并识别当前的环境。这个项目的一个关键智力优点是展示了这些新颖的神经算法如何在多个感觉域执行准确、鲁棒和低功耗的感觉分析。就像大脑能够处理来自不同传感器的数据一样。研究和开发这些神经算法将为人类大脑如何学习识别重要的感官信息，如何能够整合来自不同感官模式的信息，以及如何能够如此有效地执行复杂的分析提供见解。

The broader impact/commercial potential of this project is to enable continuous sensing applications in a wide range of energy-constrained sensor-enabled devices. Without dramatic innovations in the development of ultralow power sensory processing, continuous and accurate sensing will remain a niche application limited to environments with a stable and plentiful power source and significant computing resources. The technology described in this proposal will demonstrate the viability and potential widespread deployment of continuous sensing devices in mobile or remote environments with strict energy constraints. An important immediate market for the proposed technology with significant customer base is the smartphone and wearables market, where many new and emerging end user applications could leverage environmental sensing to trigger context-based and anticipatory actions. The proposed technology is broadly applicable to a number of other markets and domains, including medical, health, and safety monitoring of critical patient sensors, personal fitness devices, military applications, and environmental monitors. The ability to flexibly deploy continuous sensing for these and other applications has the potential to revolutionize these markets and create entirely new and unforeseen application domains. This Small Business Innovation Research Phase 2 Project plans to research and develop algorithms based on the properties of biological spiking neurons and the sensory processing capabilities of the human brain. The human brain is truly unique in its ability to use a basic computational element, the spiking neuron, and perform a broad variety of tasks. The brain has the ability to accurately classify sensory patterns from multiple modalities (touch, sight, etc.), to interpret the outside world, and to recognize the current context. A key intellectual merit of this project is a demonstration of how these novel neural algorithms can perform accurate, robust, and low power sensory analysis across multiple sensory domains. Just as the brain is capable of processing data from very different sensors. Researching and developing these neural algorithms will provide insight as to how the human brain learns to recognize important sensory information, how it is able to integrate information from such different sense modalities, and how it is able to perform complex analysis so efficiently.