CCF: Small: Smart Continually-aware High-Fidelity Sensor Interfaces

本研究解决了持续感知高保真感测的挑战。关键的挑战包括传感器接口的能耗和需要通过传感器通信的数据量，考虑到传感器必须以高保真度持续运行的两种传感场景。在第一种情况下，传感器连续处理信号并提供输出，例如，助听器连续工作，并执行复杂的高保真音频信号处理，但要长时间使用小电池。在第二种场景中，该设备检测并作用于特定的感官刺激，例如，检测和识别说话者（人）或单词或句子。这项研究的广泛影响将通过在推广和教育方面的大量努力得到加强。这些措施包括向K-12学生伸出援手，为来自未被充分代表的少数民族的本科生提供研究经验，以及更新研究生和本科课程。更具体地说，本研究探讨了一种新型的传感器接口电路，有效地实现了这两种高保真的传感场景。它通过在传感器接口架构中嵌入复杂的处理和智能来实现这一点。这种方法不同于传统的方法，传统的方法依赖于单独的信号处理设备，或者将数据发送到云端进行处理。与原有的接口电路相比，新方法的实现几乎没有额外的成本和能耗。预期的好处是整体能耗大大降低，传感器产生的数据量将减少几个数量级。

This research addresses the challenges of continually-aware high-fidelity sensing. The key challenges include the energy consumption of the sensor interface and the amount of data that needs to be communicated by the sensor by considering two sensing scenarios in which a sensor must be constantly operate with high fidelity. In the first case, the sensor continuously processes a signal and delivers an output, e.g., a hearing aid operating continuously, and performing sophisticated high-fidelity audio signal processing and yet running from a small battery for a long period of time. In the second scenario, the device detects and acts on a particular sensory stimulus, e.g., the detection and recognition of a speaker (person) or words or sentences. The broader impacts of this research will be enhanced with substantial efforts in outreach and education. These include outreach to K-12 students, research experience for undergraduates from underrepresented minorities, and updated graduate and undergraduate curriculum.More specifically, this research investigates a new type of sensor interface circuit that efficiently implements both of these high fidelity sensing scenarios. It does this by embedding sophisticated processing and intelligence within the sensor interface architecture. This approach is different from the conventional approach which relies on separate signal processing devices or alternatively sends data to the cloud for processing. The new approach is implemented with little extra cost and energy consumption, beyond that of the original interface circuit. The expected benefit is that the overall energy consumption is greatly reduced and the amount of data generated by the sensor is to be reduced by orders of magnitude.