This paper proposes a sensor data compression mechanism based on the amount of change of sensor input data and a power management scheme for various sensors used for the motion recognition application. The experimental results confirmed that the proposed compression mechanism and power management scheme reduced the wakeup count of the sensor hub core and the amount of data transmitted to the core by about 78% compared to the conventional data buffering structure, and the power consumption of the IMU (inertial measurement unit) is reduced by about 56%.

This paper proposes an adaptive cache replacement policy to select a victim block based on the reuse characteristics of stored blocks by utilizing the fine-grain reusability monitor for each cache set. The evaluation result shows that the proposed mechanism can achieve a performance improvement of about 13% on average over conventional 2 MB cache and can deliver performance comparable to four times bigger cache, i.e., an 8 MB last level cache (LLC).

No abstract

With the advent of the machine learning and IoT, many low-power edge devices, such as wearable devices with various sensors, are used for machine learning based intelligent applications, such as healthcare or motion recognition. While these applications are becoming more complex in order to provide high-quality services, the performance of conventional low-power edge devices with extremely limited hardware resources is insufficient to support the emerging intelligent applications. We designed a hardware accelerator, called an Intelligence Boost Engine (IBE), for low-power smart edge devices in order to enable the real-time processing of emerging intelligent applications with energy-efficiency and limited programmability. The measurement results confirm that the proposed IBE can reduce the power consumption of the edge node device by 75% and achieve performance improvement in processing the kernel operations of applications such as motion recognition by 69.9 times.

No abstract