Abstract: The present invention is a data cleaning device executing a data cleaning method. The data cleaning method includes executing an application, and executing a prompt plugin management program. The operation of executing the prompt plugin management program includes generating a prompt instruction based on an unformatted content via a first prompt template, transmitting the prompt instruction to a first device, and receiving a formatted content from the first device. The formatted content is generated by the first device based on the prompt instruction through a large language model.

Abstract: Provided is a sensing and adaptation method for exercise. The method is applied to a sensing and adaptation device for exercise and includes the following steps: generating an exercise game and a first resistance setting to an exercise equipment according to a first exercise target data; receiving an exercise interactive data of a user operating the exercise game from the exercise equipment; generating a reaction-time data according to a stage data of the exercise game and the exercise interactive data and generating an operation trajectory data according to the first resistance setting, the stage data and the exercise interactive data; calculating a deviation degree of the operation trajectory data; calculating a second exercise target data according to the deviation degree; generating a second resistance setting and updating the exercise game according to the second exercise target, and transmitting the second resistance setting to the exercise equipment.

Abstract: A method of correcting depth of field, performed by a first MR head-mounted device, comprises: superimposing a first virtual object on a first background image to generate an superimposed image, inputting the superimposed image into a depth-of-field correction model to generate a displayed image and display the displayed image, receiving at least one user feedback signal corresponding to the displayed image, and updating a depth-of-field parameter of the depth-of-field correction model with the at least one user feedback signal for the depth-of-field correction model to generate an updated virtual object.

Abstract: An electrical appliance status analysis device and method are provided. Based on the correlation between appliance status sequence data and electricity meter sequence data, appliances are categorized into high-correlation appliances and low-correlation appliances. For high-correlation appliances, training is conducted using the first-type model directly based on the appliance status sequence data and electricity meter sequence data. For low-correlation appliances, features are extracted to produce appliance status feature data and electricity meter feature data, which are then combined with user information feature data to train a second-type model. This method remains applicable at low sampling frequencies. Even when the data sampling rate is below 1 Hz, the resulting appliance status analysis models provide accurate analysis results, addressing the limitations of NILM technology in analyzing data with low sampling frequency.

Abstract: An electricity consumption pattern classification method involves, initially, reading multiple training electricity consumption data sets. After which, the method includes performing a first clustering phase with a first machine learning clustering algorithm according to the electricity consumption characteristics of the training electricity consumption data sets, and generating multiple first-level data groups. Then, second-level feature values of the training electricity consumption data sets are generated with a feature extraction algorithm, and a second clustering phase with a second machine learning clustering algorithm is performed to generate second-level data groups under the first-level data groups. The classification result of an unclassified data set is determined according to the average similarity between the unclassified data set and each second-level data group.

Abstract: An object tracking method includes: extracting historical moving traces corresponding to historical objects from historical images, predicting predicted locations and predicted object boxes of the historical objects, determining the historical objects are in a static state or a moving state according to a heat map, wherein the heat map is generated according to the historical images, extracting current bounding boxes corresponding to current objects from current images, comparing and calculating similarity values between the predicted object boxes and the current bounding boxes respectively, corresponding one of the current objects to one of the historical objects when the similarity value is higher than a threshold value, generating a labelled object box, and using the labelled object box to update the heat map and at least one of the historical moving traces, wherein the labelled object box is in the static state or the moving state.

Abstract: An electronic device is disclosed. The electronic device includes an inertial measurement element and a processor. The inertial measurement element is configured to obtain a first inertial measurement data when the electronic device moves. The processor is configured to perform the following operations: establishing a mixed reality environment coordinate system in correspondence according to the real space, and calculating a starting coordinate point of the electronic device in the mixed reality environment coordinate system; converting the first inertial measurement data into a first movement vector in the mixed reality environment coordinate system according to an inertial measurement data mapping model; calculating a first spatial operation point in the mixed reality environment coordinate system according to the starting coordinate point and the first movement vector in the mixed reality environment coordinate system; and executing a 3D scene editing program with the first spatial operation point.

Abstract: An augmented reality interaction system applied to a physical scene and comprising a server and a plurality of mobile devices is provided. The server stores a point cloud map corresponding to the physical scene, and one of the mobile devices uploads a physical image, role state variation data and local variation data to the server. The server compares the physical image with the point cloud map to generate orientation data of the mobile device in real time, and adjusts role data corresponding to a user according to the role state variation data and the local variation data. The server pushes the orientation data of the mobile device and the role data to the other mobile devices such that augmented reality images displayed by the other mobile devices are adjusted in real time according to the orientation data of the mobile device and the role data.

Abstract: A collision warning system and method are provided. The system includes an image capturing apparatus, a sensing apparatus, and a computing apparatus. The computing apparatus analyzes a vehicle side image to generate a plurality of objects and an object coordinate corresponding to each of the objects. The computing apparatus calculates an aerial view coordinate corresponding to each of the objects based on the object coordinates and a pitch angle corresponding to the image capturing apparatus. The computing apparatus calculates a predicted collision time between the vehicle and each of the objects based on a vehicle speed, a turning angle, and the aerial view coordinates. The computing apparatus generates a plurality of warning regions and determines a warning scope corresponding to each of the warning regions based on the predicted collision times.

Abstract: The present invention provides an encryption determining method. The method includes the following steps: receiving a side channel signal; generating a filtered side channel signal by filtering noise within the side channel signal; generating a phasor signal by utilizing a filter to covert the filtered side channel signal; locating the encrypted segment by calculating a periodicity of the phasor signal utilizing a standard deviation window; extracting at least one encrypted characteristic from the encrypted segment; and generating an encryption analytic result by recognizing the at least one encrypted characteristic according to a characteristic recognition model; wherein the encryption analytic result includes a position of the encrypted segment within the side channel signal, and an encryption type corresponding to the side channel signal. The present invention is able to automatically and efficiently locate the encryption segment and analyze the encryption type corresponding to the side channel signal.

Abstract: An information security testing method and an information security testing system of an open radio access network (O-RAN) base station are provided. The information security testing method includes at least the following steps: configuring a processing device to execute a radio resource control (RRC) signaling integrity test process to determine whether or not an RRC signaling sent by a central unit (CU) is integrity protected; and in response to determining that the RRC signaling sent by the CU is integrity protected, configuring the processing device to execute an RRC signaling security test process.

Abstract: A method for testing a core network function entity in a mobile communication system is disclosed. The method includes: performing a first test case on the core network function entity; and directly performing an integrated test with plural second test cases on the core network function entity without resetting the core network function entity in a condition where the core network function entity passes the first test case.

Abstract: An image synchronization method for multiple cameras, comprising the following steps: receiving a first video of a first camera and a second video of a second camera; capturing a first object in the first video and a second object in the second video; determining whether the first object is the same as the second object; if yes, transferring a first coordinate of the first object and a second coordinate of the second object to a uniform coordinate; regulating a timing sequence of the second video to calculate a plurality of multi-object tracking accuracy values for the second video and the first video and identifying a maximum multi-object tracking accuracy value; generating a time compensation value according to a time different corresponding to the maximum multi-object tracking accuracy value and synchronizing the first camera and the second camera according to the time compensation.

Abstract: A multimodal method for detecting video includes following step of: receiving a message to be detected to obtain a multimodal association result, which message to be detected corresponds to a video to be detected; generating a plurality of detecting conditions according to multimodal association result; searching a plurality of videos in a video detection database to obtain a target video in videos according to detecting conditions, which each of videos includes a plurality of video paragraphs respectively, which each of video paragraphs includes a piece of multimodal related data respectively; comparing detecting conditions and piece of multimodal related data of video paragraphs to obtain a matching video paragraph and using video corresponding to matching video paragraph as the target video; and outputting the target video and the video to be detected to a display device for display.

Abstract: A method for processing breast tomosynthesis images includes: obtaining multiple breast images, performing binarization operation and pooling operation to at least one marked breast image to obtain a lesion mask; training a convolutional neural network (CNN) according to the breast images and the lesion mask; obtaining multiple cropped images from the breast images of different layers according to a particular two-dimensional position, and inputting each cropped image into the CNN to obtain cropped heat maps; inputting the cropped heat maps into a recurrent neural network (RNN) in an order of the levels to obtain output heat maps, and combining the output heat maps to obtain a integrated heat map, and training the RNN according to the integrated heat map and the lesion mask.

Abstract: A method of deploying microservice includes: determining whether a current load of a task queue of a target edge host is not smaller than a load alert level; if the current load is not smaller than the load alert level, calculating a task migration number of a first queue of the task queue to deploy a microservice corresponding to the first queue at at least one of a first available edge host and a cloud host; if the current load is smaller than the load alert level, calculating a long-term load according to a history pushing rate, a history consumption rate and a default time period; and when a sum of the long-term load and a current load of a second queue of the task queue is not smaller than the load alert level, deploying a microservice corresponding to the second queue at a second available edge host.

Abstract: An RF fingerprint signal processing device configured for executing a machine learning algorithm on a plurality of input signals. The RF fingerprint signal processing device includes a receiver-feature determination circuit and a classifying determination circuit. The receiver-feature determination circuit is configured to compute on the plurality of input signals in a neural network. The classifying determination circuit is coupled with the receiver-feature determination circuit, and the classifying determination circuit is configured to send feedback information of a receiver-feature component to the receiver-feature determination circuit. The receiver-feature determination circuit decreases the receiver-feature weight of the neural network. The receiver-feature weight is associated with the receiver-feature component, and the receiver-feature weight which is decreased is applied for computing an output value of the neural network.

Abstract: A data processing method based on Trojan circuit detection includes controlling a processor, in a testing stage, to perform following steps: obtaining a plurality of characteristic values corresponding to a logic gate circuit; performing a distribution adjustment operation on the characteristic values to generate a plurality of adjusted characteristic values; and performing classification on the adjusted characteristic values to generate a logic identification result.

Abstract: A landmark identification and marking system for a panoramic image is provided. The system includes a storage device and a back-end processor. The storage device stores an initial panoramic image, attitude information, motion tracking information, and a landmark list. The back-end processor performs steps of: adjusting a visual angle of the initial panoramic image to a designated angle according to a difference value between the visual angle and the designated angle; providing the adjusted initial panoramic image to a front-end processor for calculating and generating a panoramic image integrated with landmark objects in the virtual space.

Abstract: A virtual machine backup method, performed by a first host, includes: capturing a request to write data from a virtual machine to a hard disk image file, wherein the request includes written data and input and output location information, copying the written data to a temporary storage area, calculating a first key of the written data, storing the first key, the input and output location information into a first resource location structure, pausing an operation of the virtual machine and generating a second resource location structure according to the first resource location structure, the first key and a second key, and outputting a backup data set to a second host according to the second resource location structure, wherein the backup data set includes the second resource location structure and only one of existing data and the written data when the first key and the second key are the same.