One important aspect when choosing a Bluetooth Low Energy (BLE) solution is to analyze its energy consumption for various connection parameters and desired throughput to build an optimal low-power Internet-of-Things (IoT) application and to extend the battery life. In this paper, energy consumption and data throughput for various BLE versions are studied. We have tested the effect of connection interval on the throughput and compared power efficiency relating to throughput for various BLE versions and different transactions. The presented results reveal that shorter connection intervals increase throughput for read/write transactions, but that is not the case for the notify and read/write without response transactions. Furthermore, for each BLE version, the energy consumption is mainly dependable on the data volume. The obtained results provide a design guideline for implementing an optimal BLE IoT application.
This article presents an enhanced-performance, hardware-efficient Softmax Function (SF) for a deep neural network accelerator. Softmax is used in the classification layer in deep learning models and is also used in hidden layers of advanced neural networks like Transformer and Capsule networks. The major challenge in designing efficient hardware architecture of SF is complex exponential and division computational sub-blocks. Utilizing mutual exclusivity of CO-ordinate Rotational DIgital Computer (CORDIC) algorithm, hardware-optimized pipelined CORDIC-based architecture is considered for the area, power, and enhanced throughput design. In order to maintain good accuracy in the deep learning models, the proposed SF design undergoes a Pareto study on the variation of accuracy for the number of pipeline stages. The proposed design is quantized to 16-bit precision, and inference accuracy is validated for various datasets. The SF is prototyped using Xilinx Zynq FPGA and can be operated at 685MHz. Also, ASIC implementation is performed for 45nm technology node at 5GHz of maximum operating frequency. The design achieves a validation accuracy loss of less than 2% in the account of reduced silicon area and Energy-Delay-Product(EDP) (by 12×). Post synthesis simulation result illustrates that the proposed design achieves 3× better performance in terms of area, power, and logic delay compared to state-of-the-art architectures.
Keyword spotting is an important part of modern speech recognition pipelines. Typical contemporary keyword-spotting systems are based on Mel-Frequency Cepstral Coefficient (MFCC) audio features, which are relatively complex to compute. Considering the always-on nature of many keyword-spotting systems, it is prudent to optimize this part of the detection pipeline. We explore the simplifications of the MFCC audio features and derive a simplified version that can be more easily used in embedded applications. Additionally, we implement a hardware generator that generates an appropriate hardware pipeline for the simplified audio feature extraction. Using Chisel4ml framework, we integrate hardware generators into Python-based Keras framework, which facilitates the training process of the machine learning models using our simplified audio features.
In this paper we present the generalization of the test sequencing problem, originally defined for symmetrical tests, that also covers asymmetrical tests. We prove that the same heuristics that has been employed in the traditional solution of the problem (e.g., the AO* algorithm with heuristics based on Huffman's coding) can be employed also for the generalized case. Examples are given to illustrate the approach.
Airborne Light Detection and Ranging (LIDAR) has become a mainstream technology for terrain data acquisition and mapping. High sampling density of LIDAR enables the acquisition of high details of the terrain, but on the other hand, it results in a vast amount of gathered data, which requires huge storage space as well as substantial processing effort. The data are usually stored in the LAS format which has become the de facto standard for LIDAR data storage and exchange. In the paper, a hardware accelerated compression of LIDAR data is presented. The compression and decompression of LIDAR data is performed by a dedicated FPGA-based circuit and interfaced to the computer via a PCI-E general bus. The hardware compressor consists of three modules: LIDAR data predictor, variable length coder, and arithmetic coder. Hardware compression is considerably faster than software compression, while it also alleviates the processor load.