An On-Device Learning System for Estimating Liquid Consumption from Consumer-Grade Water Bottles and Its Evaluation.

A lightweight on-device liquid consumption estimation system involving an energy-aware machine learning algorithm is developed in this work. This system consists of two separate on-device neural network models that carry out liquid consumption estimation with the result of two tasks: the detection of sip from gestures with which the bottle is handled by its user and the detection of first sips after a bottle refill. This predictive volume estimation framework incorporates a self-correction mechanism that can minimize the error after each bottle fill-up cycle, which makes the system robust to errors from the sip classification module. In this paper, a detailed characterization of sip detection is performed to understand the accuracy-complexity tradeoffs by developing and implementing a variety of different ML models with varying complexities. The maximum energy consumed by the entire framework is around 119 mJ during a maximum computation time of 300 µs. The energy consumption and computation times of the proposed framework is suitable for implementation in low-power embedded hardware that can be incorporated in consumer grade water bottles.

Calibration and Testing of Small High-Resolution Transition Edge Sensor Microcalorimeters with Optical Photons.

Pulses of narrow line-width optical photons can be used to calibrate and test sub-2 eV full-width at halfmaximum (FWHM) energy resolution transition-edge sensor (TES) microcalorimeters at low energies (< 1 keV), where it is very challenging to obtain X-ray calibration lines comparable to (or narrower than) the detector resolution. This scheme depends on the ability to resolve the number of 3 eV photons in each pulse, which we have recently demonstrated up to photon numbers of about 300. At LTD-18 we showed preliminary results obtained with this technique on a 0.25 eV baseline resolution TES microcalorimeter designed for the ultra-high-resolution subarray of the Lynx mission. The line-shape was well described by a simple Gaussian. However, the difficulty of delivering photons to the small 46 µm square absorbers resulted in a large thermal crosstalk signal, whose random nature is expected to rapidly degrade the observed energy resolution towards higher photon numbers/energies. We have since improved the coupling between the optical fiber and the TES absorber and report here our current results.