As Global Positioning System (GPS) cannot provide satisfying performance in indoor environments, indoor positioning technology, which utilizes indoor wireless signals instead of GPS signals, has grown rapidly in recent years. Meanwhile, visible light communication (VLC) using light devices such as light emitting diodes (LEDs) has been deemed to be a promising candidate in the heterogeneous wireless networks that may collaborate with radio frequencies (RF) wireless networks. In particular, light-fidelity has a great potential for deployment in future indoor environments because of its high throughput and security advantages. This paper provides a comprehensive study of a novel positioning technology based on visible white LED lights, which has attracted much attention from both academia and industry. The essential characteristics and principles of this system are deeply discussed, and relevant positioning algorithms and designs are classified and elaborated. This paper undertakes a thorough investigation into current LED-based indoor positioning systems and compares their performance through many aspects, such as test environment, accuracy, and cost. It presents indoor hybrid positioning systems among VLC and other systems (e.g., inertial sensors and RF systems). We also review and classify outdoor VLC positioning applications for the first time. Finally, this paper surveys major advances as well as open issues, challenges, and future research directions in VLC positioning systems.

Indoor wireless localization using Bluetooth Low Energy (BLE) beacons has attracted considerable attention after the release of the BLE protocol. In this paper, we propose an algorithm that uses the combination of channel-separate polynomial regression model (PRM), channel-separate fingerprinting (FP), outlier detection and extended Kalman filtering (EKF) for smartphone-based indoor localization with BLE beacons. The proposed algorithm uses FP and PRM to estimate the target’s location and the distances between the target and BLE beacons respectively. We compare the performance of distance estimation that uses separate PRM for three advertisement channels (i.e., the separate strategy) with that use an aggregate PRM generated through the combination of information from all channels (i.e., the aggregate strategy). The performance of FP-based location estimation results of the separate strategy and the aggregate strategy are also compared. It was found that the separate strategy can provide higher accuracy; thus, it is preferred to adopt PRM and FP for each BLE advertisement channel separately. Furthermore, to enhance the robustness of the algorithm, a two-level outlier detection mechanism is designed. Distance and location estimates obtained from PRM and FP are passed to the first outlier detection to generate improved distance estimates for the EKF. After the EKF process, the second outlier detection algorithm based on statistical testing is further performed to remove the outliers. The proposed algorithm was evaluated by various field experiments. Results show that the proposed algorithm achieved the accuracy of <2.56 m at 90% of the time with dense deployment of BLE beacons (1 beacon per 9 m), which performs 35.82% better than <3.99 m from the Propagation Model (PM) + EKF algorithm and 15.77% more accurate than <3.04 m from the FP + EKF algorithm. With sparse deployment (1 beacon per 18 m), the proposed algorithm achieves the accuracies of <3.88 m at 90% of the time, which performs 49.58% more accurate than <8.00 m from the PM + EKF algorithm and 21.41% better than <4.94 m from the FP + EKF algorithm. Therefore, the proposed algorithm is especially useful to improve the localization accuracy in environments with sparse beacon deployment.

We propose a sub-μW always-ON keyword spotting (μKWS) chip for audio wake-up systems. It is mainly composed of a neural network (NN) and a feature extraction (FE) circuit. For significantly reducing the memory footprint and computational load, four techniques are used to achieve ultra-low-power consumption: 1) a serial-FFT-based Mel-frequency cepstrum coefficient circuit is designed for FE, instead of the common parallel FFT. 2) A small-sized binarized depthwise separable convolutional NN (DSCNN) is designed as the classifier. 3) A framewise incremental computation technique is devised in contrast to the conventional whole-word processing. 4) Reduced computation allows a low system clock frequency, which enables near-threshold voltage operation, and low leakage memory blocks are designed to minimize the leakage power. Implemented in 28-nm CMOS technology, this μKWS consumes 0.51 μW at a 40-kHz frequency and a 0.41-V supply, with an area of 0.23 mm². Using the Google speech command data set, 97.3% accuracy is reached for a one-word KWS task and 94.6% for a two-word task.

Location service is fundamental to many IoT applications such as smart home, wearables, smart city, and connected health. With existing infrastructures, wireless positioning is widely used to provide the location service. However, wireless positioning has the limitations such as highly depending on the distribution of access points (APs); providing a low samplerate and noisy solution; requiring extensive labor costs to build databases; and having unstable RSS values in indoor environments. To reduce these limitations, this paper proposes an innovative integrated platform for indoor localization by integrating low-cost MEMS sensors and wireless signals. This proposed platform consists of wireless AP localization engine and sensor fusion engine, which is suitable for both dense and sparse deployments of wireless APs. The proposed platform can automatically generate wireless databases for positioning, and provide a positioning solution even in the area with only one observed wireless AP, where the traditional trilateration method cannot work. This integration platform can integrate different kinds of wireless APs together for indoor localization (e.g., WiFi, BLE, and RFID). The platform fuses all of these wireless distances with low-cost MEMS sensors to provide a robust localization solution. A multi-level quality control mechanism is utilized to remove noisy RSS measurements from wireless APs and to further improve the localization accuracy. Preliminary experiments show the proposed integration platform can achieve the average accuracy of 3.30 m with the sparse deployment of wireless APs (1 AP per 800 m²).

In-memory computing (IMC) is an effective solution for energy-efficient artificial intelligence applications. Analog IMC amortizes the power consumption of multiple sensing amplifiers with an analog-to-digital converter (ADC) and simultaneously completes the calculation of multi-line data with a high parallelism degree. Based on a universal one-transistor one-magnetic tunnel junction (MTJ) spin transfer torque magnetic RAM (STT-MRAM) cell, this paper demonstrates a novel tunneling magnetoresistance (TMR) ratio magnifying method to realize analog IMC. Previous concerns including low TMR ratio and analog calculation nonlinearity are addressed using device-circuit interaction. The TMR is magnified $7500\times $ using a latch structure in combination with the device. Peripheral circuits are minimally modified to enable in-memory matrix-vector multiplication. A current mirror with a feedback structure is implemented to enhance analog computing linearity and calculation accuracy. The proposed design maximumly supports 1024 2-bit input and 1-bit weight multiply-and-accumulate (MAC) computations simultaneously. The proposal is simulated using the 28-nm CMOS process and MTJ compact model. The integral nonlinearity is reduced by 57.6% compared with the conventional structure. 9.47-25.4 TOPS/W is realized with 2-bit input, 1-bit weight, and 4-bit output convolution neural network (CNN).

This paper proposes a high power-performance-area efficient background noise aware keyword-spotting (KWS) processor based on an optimized binarized weight network (BWN). To reduce the power consumption while maintaining the system recognition accuracy for different background noise, the KWS processor with a SNR prediction module can be adaptively configured to use dual computing modes (standard computing mode and approximate computing mode) for both high recognition accuracy under high background noise and ultra-low power consumption under low background noise. The mel-scale frequency cepstral coefficients (MFCC) module is optimized with approximate computing technologies, which can reduce the power consumption by up to 3.1x and 5.7x for high/low background noise, respectively. Based on the evaluation of the architecture design space exploration, an ultra-low power BWN accelerator with low voltage, area and leakage power and using precision self-adaptive approximate computing units was proposed. Evaluated under 22nm process technology, this work can support up to 10 keywords real time recognition with power consumption of 15.1μW for high background noise and 10.8μW for low background noise. Compared to the state-of-the-art KWS architectures, our work can achieve ultra-low power consumption (about 1.7x reduced), while maintaining high system capability and adaptability.