Energy efficient integrated MEMS neural network for simultaneous sensing and computing

Abstract

Biological systems seamlessly combine multiple functions in lightweight and energy-efficient structures. Such capability in synthetic structures would be desirable in numerous engineering applications such as aerospace, robotics and wearable devices. Here we report an integrated silicon-based structure configured to sense, perform different classification algorithms, and produce an action signal within the same physical layer. The algorithms are coded in the mechanical responses of the sensing elements of multiple coupled micro-electro-mechanical systems (MEMS), simultaneously capturing acceleration measurements to produce an actuated signal. This all-in-one structure operates with zero circuitry and low power consumption. As a demonstration, we designed and fabricated a network of three MEMS neurons to successfully perform both simple signal classification and activity recognition problems (standing and sitting) with only 9.92 × 10−17 kWh and 17.79 × 10−19 kWh energy consumption per operation, respectively. Our approach will enable emergent technologies, such as wearable devices, to perform complex computations with power from a single battery charge.

In-sensor human gait analysis with machine learning in a wearable microfabricated accelerometer

“…Lower power consumption levels could arguably be achieved by eliminating the feedback circuitry to create activations entirely in the mechanical domain. This could be achieved using multiple resonators 47 or multiple proof masses 48 . A completely mechanical implementation with the number of activations (approximately 100) and memory (on timescales on the order of seconds) required for complex time-series processing (such as gait analysis) has yet to be experimentally demonstrated, but hybrid systems with a few resonators or proof masses 25 , 49 could be a stepping stone toward this goal, that proportionally reduces the power consumption of the feedback electronics.…”

MEMS reservoir computing system with stiffness modulation for multi-scene data processing at the edge

“…Moreover, the three layers (input layer, reservoir layer, and output layer) were always set up separately at the hardware level, resulting in a discrete system with a large volume. Some improvements have been proposed, such as using bias time multiplexing to divide input and mask 13 , 14 , using hybrid nonlinearity (HNL) to enhance RC ability for classification tasks 15 , 16 , and using structural design to obtain MEMS neurons 17 , 18 . However, drawbacks need to be considered, such as feedback still existing, resulting in a separate system, poor long-term memory capacity (MC), and pending tasks have to be designed (basically simple classification tasks), especially for the using device, respectively.…”