Biomimetic Architectures for Sensing & Signal Processing

仿生传感架构

• 批准号: 2323940
• 金额: --
• 依托单位: University of Manchester
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2323940
• 关键词: Biomimetic; Architectures; Sensing; Signal; Processing

The biological systems are capable of sensing, processing and communicating, information robustly with extremely low-power consumption, using hybrid neural network architectures constructed with slow, less accurate, highly heterogeneous neural elements, neurons and synapses. Most of this sensory and neural processing performed using spikes-based computations. Investigating and understanding these inherent spikes-based computational methods used in the sensors, sensory-pathways and cortex may help us to envisage and build future technologies that are energy-efficient and intelligent. Initially, the project investigates the spike-based sensing, communicating and processing principles by mimicking biological sensory and processing architectures in modern hardware technologies. Additionally, the student will also design and implement conventional biomedical architecture to sense and process a bio-signal to minimise noise and power consumption. Then, the student will investigate the ability to incorporate the bio-inspired techniques in sensors and signal processing architectures to realise low-power biomedical sensors and circuits that are less susceptible to noise and fault tolerance or robust to variability.The work carried out involve implementation of neuromorphic hardware cochleae, using a novel spike encoding algorithm and Gammatone filters, mimicking the operation of the biological cochlea. In parallel, the student will also develop fetal ECG signal capturing and signal processing architecture using conventional electronics. Then, the student will investigate and implement bio-inspired spikes-based techniques to detect ECG and fetal ECG signals.

生物系统能够以极低的功耗健壮地感知、处理和交流信息，使用由缓慢、不准确、高度异构的神经元件、神经元和突触组成的混合神经网络架构。大部分的感觉和神经处理都是通过基于峰值的计算来完成的。研究和理解传感器、感觉通路和皮层中使用的这些基于固有尖峰的计算方法，可能有助于我们设想和构建未来节能、智能的技术。最初，该项目通过模仿现代硬件技术中的生物感官和处理架构来研究基于尖峰的传感、通信和处理原理。此外，学生还将设计和实施传统的生物医学建筑来感知和处理生物信号，以最大限度地减少噪音和功耗。然后，学生将研究在传感器和信号处理架构中结合生物启发技术的能力，以实现低功耗生物医学传感器和电路，这些传感器和电路不易受噪声和容错性的影响，或对变异性的鲁棒性。这项工作涉及到神经形态硬件耳蜗的实现，使用一种新的脉冲编码算法和伽玛酮滤波器，模拟生物耳蜗的运作。同时，学生也将开发胎儿心电图信号捕获和信号处理架构，使用传统的电子技术。然后，学生将研究和实现基于生物脉冲的技术来检测ECG和胎儿ECG信号。