AI / ML

Cheng-Han Yang, Cheng-Xin Xue, En-Yu Yang, Yen-Kai Chen, Yun-Sheng Chang, Tzu-Hsiang Hsu, Ya-Chin King, Chorng-Jung Lin, Ren-Shuo Liu, Chih-Cheng Hsieh, Kea-Tiong Tang, Meng-Fan Chang National Tsing Hua University, Hsinc

Jan M. Rabaey2, H.-S. Philip Wong1, Max M. Shulaker3, Subhasish Mitra1 Stanford University, Stanford, CA University of California, Berkeley, Berkeley, CA 3 Massachusetts Institute of Technology, Cambridge, MA 1 2 We demo

Embedded sensory systems (Fig. 31.2.1) continuously acquire and process data for inference and decision-making purposes under stringent energy constraints. These always-ON systems need to track changing data statistics a

local column that replicates the local bit-line capacitance. This process continues until the voltage of the rail being integrated exceeds the other one, at which point the SA output flips. This signals conversion comple

Narrowband Internet-of-Things (NB-IoT) is a newly developed 3GPP protocol optimized for low-power wide-area IoT applications and is evolving toward the future fifth-generation (5G) mobile communication. It specifies at l

high data-rates, the millimeter-wave (mmW) 5G communication standard will extensively use high-order complex-modulation schemes (e.g., QAM) with high peak-to-average power ratios (PAPRs) and large RF bandwidths. High-eff

Aurel A. Lazar, Mingoo Seok Columbia University, New York, NY Voice user interfaces (UIs) are highly compelling for wearable and mobile devices. They have the advantage of using compact and ultra-low-power (ULP) input de

the W-band. At 100GHz, DCOs become more sensitive to parasitics resulting in low frequency resolution. A high-resolution delta-sigma TDC is used to reduce quantization noise by noise-shaping in [2], but it suffers from l

Compressive sensing (CS) techniques enable new reduced-complexity designs for sensor nodes and help reduce overall transmission power in wireless sensor network [1-2]. Prior CS reconstruction chip designs have been descr

Lund University, Lund, Sweden data movements using holding buffers to maximize data reuse. The data reuse is especially advantageous in our design, as it requires a relatively long 28b data bit width to support a wide ra

latency, bandwidth, and privacy has created demand for low-energy deep convolutional neural networks (CNNs). The single-layer classifier in [1] achieves sub-nJ operation, but is limited to moderate accuracy on low-comple

Recently, 3D hand-gesture recognition (HGR) has become an important feature in smart mobile devices, such as head-mounted displays (HMDs) or smartphones for AR/VR applications. A 3D HGR system in Fig. 13.4.1 enables user

deep learning algorithms from face recognition to emotion recognition in mobile or embedded environments [3]. However, most works accelerate only the convolutional layers (CLs) or fully-connected layers (FCLs), and diffe

Shinya Takamaeda-Yamazaki1, Junichiro Kadomoto2, Tomoki Miyata2, Mototsugu Hamada2, Tadahiro Kuroda2, Masato Motomura1 Hokkaido University, Sapporo, Japan Keio University, Yokohama, Japan 1 2 A key consideration for deep

Cost-efficient, point-of-use diagnostics are critical for early disease detection. Traditionally, the majority of lab-based analysis equipment utilizes fluorescent markers for biodetection assays. However, magnetic-based

ConvNets, or Convolutional Neural Networks (CNN), are state-of-the-art classification algorithms, achieving near-human performance in visual recognition

Recently, deep learning with convolutional neural networks (CNNs) and recurrent neural networks (RNNs) has become universal in all-around applications. CNNs are used to support vision recognition and processing, and RNNs

Elio Guidetti1, Fabio De Ambroggi4, Tommaso Majo1, Paolo Zambotti4, Manuj Ayodhyawasi2, Harvinder Singh2, Nalin Aggarwal2 STMicroelectronics, Cornaredo, Italy STMicroelectronics, Greater Noida, India 3 STMicroelectronics

Internet-ofEverything (IoE) devices to data center servers for intelligent recognition processes is impractical for energy reasons, requiring in-situ processing of such data. However, algorithms accelerated by previous r

Nvidia, Westford, MA 1 2 Deep learning using convolutional neural networks (CNN) gives state-of-the-art accuracy on many computer vision tasks (e.g. object detection, recognition, segmentation). Convolutions account for

forward-collision warning, advanced emergency braking, adaptive cruise control, and lane-keeping assistance. Recently, automotive black boxes are installed in cars for tracking accidents or theft. In this paper, a dual-m

analysis in image retrieval with high accuracy [1]. As Fig. 4.6.1 shows, various applications, such as text, 2D image and motion recognition use DL due to its best-in-class recognition accuracy. There are 2 types of DL:

that integrates an image sensor and parallel image processors on a single silicon die. Nowadays, high-speed vision chips with powerful recognition capabilities are greatly demanded in applications such as: industrial aut

Modern high-data-rate communication systems such as LTE use spectrally efficient modulation schemes with a high peak-to-average power ratio (PAPR), placing stringent linearity demands on the RF power amplifiers (PA). The

To optimize the power consumption and system performance of battery-supplied devices, it is required to monitor and adjust the transmitted RF power accurately and continuously. This is typically done by an external power

augmentation of images in a real-world environment. Wearable systems, such as head-mounted display (HMD) systems, have attempted to support real-time AR as a next generation UI/UX [1-2], but have failed, due to their lim

(µEHs) is increasing (for seamless energy source in applications such as wireless sensor node), two major problems still obstruct versatile use of them. The first problem is the self-startup capability. Because many wire

techniques has been developed to satisfy various electrical and electronics applications requirements [1,2]. In high-voltage applications, the main issue is the electrical isolation with accurate measurement at low signa

Intel, Hillsboro, OR As CMOS technology continues to scale for SoC applications, significant challenges to implement a monolithic linear high-power amplifier have emerged. This results from the low breakdown voltage of t

challenges: (1) the large number of features to process requires high computational power, and (2) false matches from background clutter can degrade recognition accuracy. Previously, saliency based bottom-up visual atten

The opening up of the mm-wave band has created opportunities for high-data-rate communication, radar and medical imaging. The cost and size advantages of CMOS have motivated research on 60GHz CMOS front-end design [1]. H