The rapid growth of artificial intelligence and the bottleneck computing force development have promoted emergence new architectures. In-memory computing, as a structure that breaks traditional separation memory is considered to be method break force. This paper presents two typical parts architecture design for in-memory IM-A IM-P. And it expounds implementations in various storage media. challenges manufacture, design, application are summarized. trend mainly realize algorithms.

Motion capture system has been widely used in virtual reality and rehabilitation area. This study proposed a data-driven method using low-dimensional input of inertial motion to reconstruct human lower-limb motions. The long short-term memory (LSTM) neural network was an ensemble LSTM architecture involved improve reconstruction performance. Besides, the selection optimal sensor configuration scheme time-step parameters discussed detail. experiment shows that could get lowest joint angle...

Motion capture system serves as a critical technology in wide range of applications. Nowadays, motion based on inter components has become new research hotspot. This paper provides design low cost wearable wireless inertia measurement unit 9-DOF micro sensor MPU9250 and STM32F103C8T6 MCU. Furthermore, an ultra-low power (ULP) 2Mbps RF transceiver IC NRF24L01 was used transceiver, which operates 2.4GHz ISM (Industrial, Scientific Medical) band, with peak RX/TX currents lower than 14mA....

Gait phase detection is of great significance in the field motion analysis and exoskeleton-assisted walking, can realize accurate control exoskeleton robots. Therefore, order to obtain gait information ensure good accuracy, a recognition framework based on New Hidden Markov Model (NHMM) proposed improve accuracy detection. A multi-sensor data acquisition system was developed used collect training eight healthy subjects measure acceleration plantar pressure human body. Accuracy framework,...

Hypergraph neural networks (HGNN) have shown superior performance in various deep learning tasks, leveraging the high-order representation ability to formulate complex correlations among data by connecting two or more nodes through hyperedge modeling. Despite well-studied adversarial attacks on Graph Neural Networks (GNN), there is few study against HGNN, which leads a threat safety of HGNN applications. In this paper, we introduce HyperAttack, first white-box attack framework hypergraph...

Gait symmetry training plays an essential role in the rehabilitation of hemiplegic patients. Robotics-based gait has been widely accepted by patients and clinicians. Reference trajectory generation for affected side using motion data unaffected is important way to achieve this. However, online reference requires algorithm provide correct phase delay could reduce impact measurement noise from sensors input uncertainty users. Based on active knee orthosis (AKO) prototype, this work presents...

Edge AI is an emerging paradigm that leverages edge computing to pave the last mile delivery of artificial intelligence. To satisfy stringent timeliness and energy-efficiency requirements tasks, specialized accelerator Neural Processing Units (NPU) have been widely equipped by nodes. Compared traditional centralized processing units (CPU), NPU has better performance energy-efficiency. However, these benefits come at cost reduced inference accuracy. As a result, existing coarse-grained...

This work proposes a rehabilitation training assessment method combined with virtual reality technology (VR) to improve the effect of stroke patients( upper limb rehabilitation. During training, patient(s location information and joint angle features are collected by Kinect2.0 somatosensory equipment. The data is filtered median filtering algorithm. Then modified dynamic time warping (DTW) algorithm used recognize motion. Finally, system considers both influence duration amplitude in result...

Spatio-temporal action detection (STAD) is a task receiving widespread attention and has numerous application scenarios, such as video surveillance smart education. Current studies follow localization-based two-stage paradigm, which exploits person detector for localization feature processing model with classifier classification. However, many issues occur due to the imbalance between settings complexity in STAD. Firstly, of heavy offline detectors adds inference overhead. Secondly,...

Deep Hashing is widely used for large-scale image-retrieval tasks to speed up the retrieval process. Current deep hashing methods are mainly based on Convolutional Neural Network (CNN) or Vision Transformer (VIT). They only use local global features low-dimensional mapping and similarity loss function optimize correlation between pairwise triplet images. Therefore, effectiveness of limited. In this paper, we propose a dual-stream correlation-enhanced framework (DSCEH), which uses image...

Customized human-machine interfaces for controlling assistive devices are vital in improving the self-help ability of upper limb amputees and tetraplegic patients. Given that most them possess residual shoulder mobility, using it to generate commands operate can serve as a complementary approach brain-computer interfaces.

Approach: We propose hybrid body-machine interface prototype integrates soft sensors an inertial measurement unit. This study introduces both rule-based data...

With the development of Internet technology, more and businesses are developing towards microservices. As entrance request traffic, API gateway carries a huge load. It needs to avoid becoming bottleneck entire system have high performance. At same time, traditional implementation method strongly relies on processing power CPU itself. However, with slowdown Moore's Law sharp increase business types data volume, CPU's performance improvement gateways is limited. In response these problems,...

The real-time gait phase of human lower extremity is the foundation for wearable robots to provide precise and complex assistance strategies in human-robot interaction. In addition strengths estimation performance, it crucial make devices portable user-friendly that can drive adoption unstructured environments. this paper, we present an online continuous system based on multi-source flexible sensors address issue. Specifically, utilize two soft bend mounted around hip joint a set pressure...

The increasing popularity of artificial intelligence (AI) requires the ability to process intensive data and efficient heterogeneous computing power. As a result, integration scheme involving both central processing units (CPUs) neural (NPUs) has become increasingly prevalent in various edge terminals, such as mobile phones. Compared with traditional separated solutions, can effectively reduce distance number transmissions, thereby accelerating deep network (DNN) models improving energy...