NeuroVE: Brain-Inspired Linear-Angular Velocity Estimation With Spiking Neural Networks

Online leakage current classification using convolutional neural network long short-term memory for high voltage insulators on web-based service

Associations Between Key Language-Related Measures in Typically Developing School-Age Children

With the rapid development of biomedical research and information technology, the number of clinical medical literature has increased exponentially. At present, COVID‐19 clinical text research has some problems, such as lack of corpus and poor annotation quality. In clinical medical literature, there are many medical related semantic relationships between entities. After the task of entity recognition, how to further extract the relationships between entities efficiently and accurately becomes very critical. In this study, a COVID‐19 clinical trial data relationship extraction model based on deep learning method is proposed. The model adopts MPNet model, bidirectional‐GRU (BiGRU) network, MAtt mechanism and Conditional Random Field inference layer integration architecture and improves the problem that static word vector cannot represent ambiguity through pre‐trained language model. BiGRU network is used to replace the current Bi directional long short term memory structure and simplify the network structure of Long Short Term Memory to improve the training efficiency of the model. Through comparative experiments, the proposed method performs well in the COVID‐19 clinical text entity relation extraction task.

In industrial production, the characteristics of compressor vibration signal change with the production environment and other external factors. Therefore, to ensure the effectiveness of the model, the vibration signal prediction model needs to be updated constantly. Due to the complex structure of Long Short Term Memory (LSTM) network, the LSTM model is difficult to adapt to the scene of online update. Therefore, the update model based on LSTM is difficult to respond quickly to data changes, which affects the accuracy of the model. To solve this problem, the online learning algorithm is introduced into prediction model, Error-LSTM (E-LSTM) model is proposed in this paper. The main idea of E-LSTM model is to improve the accuracy and efficiency of the model according to test error of the model. First, the hidden layer neurons of LSTM network are divided into blocks, and only part of the modules are activated at each time step. The number of modules activated is determined by test error. Thus, the training speed of the model is accelerated and the efficiency of the model is improved. Second, the E-LSTM model can adaptively adjust the training method according to the data distribution characteristics, so as to improve the accuracy of updated model. In experimental part, two types of datasets are used to verify the performance of the proposed model. LSTM model is used for comparative experiments, and the results showed that the updating model based on E-LSTM is better than that based on LSTM in terms of model accuracy and efficiency.

A neural network structure of Long Short Term Memory (LSTM) is proposed which could be used to predict the temperature and humidity of other key parts from the temperature and humidity data of some parts of the human body when wearing tight sportswear, so as to realize the temperature and humidity data prediction of all key points of the human body. The temperature and humidity of different people wearing tights were collected by DHT sensors. The experimental results show that the LSTM neural network structure proposed has higher prediction accuracy than other algorithms, and the model evaluates the feasibility of temperature and humidity data of tights in a state of motion, which facilitates the study of dynamic thermal and humid comfort and reduces the time cost of analyzing the temperature and humidity distribution and changing the law during human movement. It will effectively promote the study of temperature and humidity changes when people wear sports tights, provide theoretical reference for the study of human skin temperature in the field of sports medicine, and provide practical guidance for the application of human skin temperature changes in sports clothing production, diagnosis and prevention of sports injuries.

Efficient velocity estimation plays an important role in the stability of haptic interfaces. In this work, a new digital circuit is realized to reduce the noise level in low velocity estimation. The proposed adaptive-method (A-method) is based on the concept of measuring the time-interval between two or more incoming quadrature pulses from the optical encoder to the order of picoseconds (ps). A time-to-digital converter (TDC) is used initially to implement the conventional velocity estimation techniques like frequency-count method (M-method) and period-count method (T-method). Later, the two methods are implemented simultaneously using the TDC for the adaptive estimation. The range of TDC used is 0–110 nanoseconds with 420–425 ps resolution. A carry-chain was implemented in order to remove non-linearity and increase precision. Experimental results demonstrate the superiority of the proposed A-method over the conventional T- and M-methods.

Micro Aerial Vehicles (MAVs) have become increasingly popular due to their wide applications in many areas. For autonomous navigation and safe control of MAVs, it is essential to have accurate and reliable velocity and position estimation. However, due to limited computational power and payload, it is still challenging for autonomous operation of MAVs in complex environments. In this paper, we propose a robust and efficient velocity estimation framework for MAVs with a single downward-facing RGB-D camera, which is able to provide metric velocity estimation in three dimensions as well as yaw rate in real-time without the fusion of additional sensors. Unlike traditional algorithms, our method uses kernel cross-correlators (KCCs) to efficiently determine optical flow for motion estimation, which does not rely on costly feature extraction and matching process. Moreover, we utilize depth images to estimate the vertical velocity of MAVs without the assumption of a flat ground. Autonomous flight tests on a quadrotor in complex environments demonstrate the robustness and efficiency of our method.

Spiking Neural Networks (SNNs) offer a promising avenue for developing adaptive intelligence that can operate effectively, when subject to energy constraints. This is especially true for emerging gradient-based SNNs that are designed for continuous learning. However, hardware efficiency and complex SNN functionality are at odds with each other, leading to large design space, with many possible trade-offs. This paper discusses the co-optimization of SNNs and hardware architectures. We compare digital designs and emerging compute-in-memory architectures, in terms of operation. We discuss the effect of different layer-types on memory organization and outline how different memory organization schemes are impacted by the SNN layertype. Finally, we outline some open problems that must be solved to design the next-generation of intelligent hardware systems.

Singing voice conversion is a task to convert a song sang by a source singer to the voice of a target singer. In this paper, we propose using a parallel data free, many-to-one voice conversion technique on singing voices. A phonetic posterior feature is first generated by decoding singing voices through a robust Automatic Speech Recognition Engine (ASR). Then, a trained Recurrent Neural Network (RNN) with a Deep Bidirectional Long Short Term Memory (DBLSTM) structure is used to model the mapping from person-independent content to the acoustic features of the target person. F0 and aperiodic are obtained through the original singing voice, and used with acoustic features to reconstruct the target singing voice through a vocoder. In the obtained singing voice, the targeted and sourced singers sound similar. To our knowledge, this is the first study that uses non parallel data to train a singing voice conversion system. Subjective evaluations demonstrate that the proposed method effectively converts singing voices.

Exploring diverse and fine-grained caption for video by incorporating convolutional architecture into LSTM-based model

Financial Time Series Forecasting (TSF) remains a critical challenge in Artificial Intelligence (AI) due to the inherent complexity of financial data, characterized by strong non-linearity, dynamic non-stationarity, and multi-factor coupling. To address the performance limitations of Spiking Neural Networks (SNNs) caused by hyperparameter sensitivity, this study proposes an SNN model optimized by an Improved Cuckoo Search (ICS) algorithm (termed ICS-SNN). The ICS algorithm enhances global search capability through piecewise-mapping-based population initialization and introduces a dynamic discovery probability mechanism that adaptively increases with iteration rounds, thereby balancing exploration and exploitation. Applied to futures market price difference prediction, experimental results demonstrate that ICS-SNN achieves reductions of 13.82% in MAE, 21.27% in MSE, and 15.21% in MAPE, while improving the coefficient of determination (R2) from 0.9790 to 0.9822, compared to the baseline SNN. Furthermore, ICS-SNN significantly outperforms mainstream models such as Long Short-Term Memory (LSTM) and Backpropagation (BP) networks, reducing prediction errors by 10.8% (MAE) and 34.9% (MSE), respectively, without compromising computational efficiency. This work highlights that ICS-SNN provides a biologically plausible and computationally efficient framework for complex financial TSF, bridging the gap between neuromorphic principles and real-world financial analytics. The proposed method not only reduces manual intervention in hyperparameter tuning but also offers a scalable solution for high-frequency trading and multi-modal data fusion in future research.

Predicting system faults is critical to improving productivity, reducing costs, and enforcing safety in industrial processes. Yet, traditional methodologies frequently falter due to the intricate nature of the task. This research presents a novel use of spiking neural networks (SNNs) in anticipating faults in syntactical time series, utilizing the generalized stochastic Petri net (GSPN) model. The inherent ability of SNNs to process both time and space aspects of data positions them as a prime instrument for this endeavor. A comparative evaluation with long short-term memory (LSTM) networks suggests that SNNs offer comparable robustness and performance.

Radar network topology optimization for joint target position and velocity estimation

It is true that the "best" neural network is not necessarily the one with the most "brain-like" behavior. Understanding biological intelligence, however, is a fundamental goal for several distinct disciplines. Translating our understanding of intelligence to machines is a fundamental problem in robotics. Propelled by new advancements in Neuroscience, we developed a spiking neural network (SNN) that draws from mounting experimental evidence that a number of individual neurons is associated with spatial navigation. By following the brain's structure, our model assumes no initial all-to-all connectivity, which could inhibit its translation to a neuromorphic hardware, and learns an uncharted territory by mapping its identified components into a limited number of neural representations, through spike-timing dependent plasticity (STDP). In our ongoing effort to employ a bioinspired SNN-controlled robot to real-world spatial mapping applications, we demonstrate here how an SNN may robustly control an autonomous robot in mapping and exploring an unknown environment, while compensating for its own intrinsic hardware imperfections, such as partial or total loss of visual input.

In this letter, an aggressive quadrotor ball playing system called Bat is proposed, whose goal is to intercept a flying ball and volley it towards a designated target. Aggressive means Bat operates the quadrotor aggressively to intercept balls that are far away and hit them to distant positions in ways that are beyond the reach of existing methods. The trajectory prediction of the ball is achieved by integrating forward the current position and velocity estimates using an extended kalman filter, and implementing cubic interpolation at the time resolution to calculate the continuous gradient for optimization. Facing the challenge of finding feasible hitting actions under extreme circumstances, we propose a two-stage planning approach, including transition point design and hitting primitive generation, with a simplified expression of uncoupled hitting actions. To obtain the best hitting motion, a trajectory optimization method is proposed, which can jointly optimize the hitting terminal states and time cost, considering dynamic feasibility and anticollision constraints. To avoid pathological hitting, a defensive rule constraint and its constraint transcription method are proposed. The largest difference from the existing methods is that Bat Planner can independently decide how to execute more aggressive keyvolleying maneuvers. A large number of simulation and real-world experiments are conducted, which prove the flying ball player can hit arriving balls from different directions and distances to arbitrary targets. To the best of our knowledge, Bat is currently the closest a quadrotor ball player has approached to human ball players' volleying ability.