Hong Kong Applied Science and Technology Research Institute Company Limited

A system using positioning with non-line of sight information from sensors is provided. The system includes a plurality of TRPs, a UE, and a backend server. The TRPs are distributed in a target region. The UE is placed within the target region and the UE comprises at least one sensor for gathering surrounding information at the target region. The backend server comprises a signal receiver, a positioning module, a signal emitter, and a feedback module. The signal receiver is configured to receive first SRSs. The positioning module is configured to compute a preliminary location of the UE according to the first SRS signals. The signal emitter sends a location signal containing the preliminary location of the UE to the UE. The feedback module is configured to receive second SRSs and NLOS information from the UE and to generate at least one positioning estimation result by positioning the UE.

A system includes a first FP-to-BFP converter, a second FP-to-BFP converter, an 8-bit integer multiplier, an adder, an accumulator, and a BFP-to-FP converter. The first and second FP-to-BFP converters receive 32-bit floating-point pixel and filter data, respectively, reducing their mantissas to 8-bit BFP format. The 8-bit integer multiplier processes these BFP values via multiply-accumulate operations, generating a 16-bit product. The adder accumulates multiple 16-bit products into a 64-bit sum, which the accumulator further aggregates. The BFP-to-FP converter transforms the 64-bit accumulated sum into a 32-bit floating-point output.

An apparatus for intrusion detection and prevention of cyber threat intelligence is provided. The apparatus includes a feature extraction model, a feature selection model, a flow-based intrusion detection module, a packet-based intrusion detection module, and an IPM. The feature extraction model is configured to parse packets from network raw packet data into sessions as a network flow dataset. The feature selection model is configured to select flow-based features from the network flow dataset to generate network flow meta. The flow-based intrusion detection module is configured to generate flow-based labels. The packet-based intrusion detection module is configured to generate packet-based labels. The IPM is configured to perform intrusion analysis according to the flow-based labels and the packet-based labels so as to generate IPM rules for intrusion detection and prevention, thereby identifying and processing the network raw packet data.

A method and apparatus for prompting-based rehearsal-free continual learning with dynamic prototypes-involved self-attention. A machine learning model with a base training set accepts continual new data inputs and use them to compute a prototype self-attention block which updates the model for the task associated with the new data without excessive or undesired overwriting of the existing model. Gini penalties are applied to the results of such computation to prevent excessive or undesired overwriting of the existing model.

An ultrasonic echo envelope detector has an analytic transformer receiving a digitized ultrasonic echo waveform that performs a band pass Finite-Impulse-Response (FIR) filter and a Hilbert transform to generate inline and quadrature outputs for a detected envelope. The analytic transformer has three FIR units and one muxed Hilbert/FIR unit, each with N taps, for a 4N-tap FIR filter. Each unit has a shared multiplier that selects one of the N taps to multiply with a coefficient. Products are summed by a band pass filter accumulator. After all N taps are processed, the accumulator output is shifted in to another N Hilbert taps in the muxed Hilbert/FIR unit. Each of the N Hilbert taps is multiplied by a corresponding Hilbert coefficient and accumulated. The middle Hilbert tap is delayed and output as the inline output while the Hilbert accumulator value is output as the quadrature output.

An electrophotographic photoreceptor, which includes a conductive substrate; an undercoat layer disposed on the conductive substrate and capable of conforming to the contour of the conductive substrate, a photoconductor charge generation layer disposed on the undercoat layer and capable of conforming to the contour of the undercoat layer, and a charge transport layer disposed on the photoconductor charge generation layer. The charge transport layer contains a charge transport material, a binder resin, a fluorine-containing resin, and a plurality of polyhedral oligomeric silsesquioxane (POSS) particles evenly dispersed in the binder resin, and the POSS particles are interconnected with at least one fluorine group and at least two non-fluoridated groups. By adding approximately 1% of the POSS particles and lubricant nanoparticles, a life-time improvement of at least 20% is achieved for an OPC drum.

The invention discloses a method of manufacturing an electrochemical cell having a polymer separator membrane for separation of electrodes in the electrochemical cell, including providing a cathode and providing a polymer separator membrane. At least one cycle of irradiating the polymer separator membrane is performed by an energy beam under a radiation dose ranging between 50 and 200 kGy to effect a cross-linking in the polymer separator membrane. The polymer separator membrane is maintained at a temperature between 30° C. and 70° C. An anode is then provided. Subsequently, the polymer separator membrane is compressed between the cathode and the anode. An electrolyte is provided to form the electrochemical cell.

A bidirectional circuit breaker has Breaking Speed Control (BSC) to prevent false triggering when an upstream fault occurs. A semiconductor switch has a Metal Oxide Semiconductor Field Effect Transistor (MOSFET), an Insulated Gate Bipolar Transistor (IGBT), and a varistor in three parallel branches that drive a positive output. An input capacitor between positive and negative inputs forms an LC network with an inductor connected to the positive input. A current sensor is connected between the inductor and the semiconductor switch. When the current sensor detects reverse current, a snubber switch is closed, connecting a snubber capacitor across both terminals of the inductor. A LCC network is formed that stores more energy than the LC network, spreading out a current spike caused by the upstream fault, reducing peak reverse current pulled through the circuit breaker by the upstream fault. Thus breaking speed is reduced when the snubber switch closes.

Methods and a non-transitory computer-readable storage medium for fine-grained action classification and/or regression are disclosed. The method includes: receiving a video stream capturing a sequence of human subject actions; identifying reference objects with spatial-temporal relationships to the action sequence; extracting a pose dataset representing the action sequence; extracting object datasets representing spatial positions of the reference objects; generating a compound data structure integrating the pose dataset and object datasets; and inputting the compound data structure into a trained machine learning model for classification and/or regression.

In a decision-support system for gastrointestinal (GI) endoscopy, convolutional neural networks (CNNs) are set up to perform decision-support tasks according to endoscopic images. Each learnable kernel used in the CNNs is advantageously modeled as a linear combination of a set of fixed kernels for simplifying kernel learning, giving a lightweight kernel model to advantageously reduce required computation resources. Further computation-resource reduction can be made by CNN model compression via knowledge distillation and by using multi-task CNNs. It enables the decision-support system to be realized as an edge computing system near a site of performing endoscopic examinations. The system can be automatically configured for esophagogastroduodenoscopy (EGD) or colonoscopy. In the system, lesion-detection results and quality-control results can be seamlessly integrated to provide value-added results, which are more valuable to the endoscopist than separately considering the lesion-detection results and quality-control results.