91results about How to "Lower confidence" patented technology

An ambulatory medical device including a plurality of electrodes configured to be disposed at spaced apart positions about a patient's body, an electrode signal acquisition circuit, and a monitoring circuit. The acquisition circuit has a plurality of inputs each electrically coupled to a respective electrode of the plurality of electrodes and is configured to sense a respective signal provided by a plurality of different pairings of the plurality of electrodes. The monitoring circuit is electrically coupled to an output of the acquisition circuit and is configured to analyze the respective signal provided by each of the plurality of different pairings and to instruct the acquisition circuit to select at least one of the plurality of different to pairings to monitor based on at least one of the quality of the respective signal, a phase difference between the respective signal and that of other pairings, a position of electrodes relative to the patient's body, and other criteria.

A system and method for computer vision driven applications in an environment that can include collecting image data across an environment; maintaining an environmental object graph from the image data whereby maintaining the environmental object graph is an iterative process that includes: classifying objects, tracking object locations, detecting interaction events, instantiating object associations in the environmental object graph, and updating the environmental object graph by propagating change in at least one object instance across object associations; and inspecting object state for at least one object instance in the environmental object graph and executing an action associated with the object state. The system and method can be applied to automatic checkout, inventory management, and / or other system integrations.

A method for classifying input provided to a digitizer sensor includes sampling output over one or more sampling periods, identifying a location of intentional input to the digitizer sensor from the output sampled over the one or more sampling periods, identifying an area of potential palm input based on the identified location of the intentional input, and classifying output detected in the area of the potential palm input as output potentially obtained from undesired input to the digitizer sensor. The area of potential palm input is defined to have a defined spatial relation to the location of the intentional input. The output classified is sampled over one or more sampling periods other than the one or more sampling periods from which the area of potential palm input is identified.

Systems and techniques are provided for blocked sensor detection and notification. A signal may be received from a sensor. The sensor may be determined to be producing anomalous output based on checking the signal from the sensor against a signal from a second sensor on the same sensor device as the sensor, checking the signal from the sensor against signals from sensors on additional sensor devices, and checking the signal from the sensor against a slow temporal model for the sensor. In response to the determination that the sensor is producing anomalous output, a notification may be generated indicating the sensor is producing anomalous output, and a confidence level in the sensor may be degraded.

In a confidence computing method and system, a processor may interpret speech signals as a text string or directly receive a text string as input, generate a syntactical parse tree representing the interpreted string and including a plurality of sub-trees which each represents a corresponding section of the interpreted text string, determine for each sub-tree whether the sub-tree is accurate, obtain replacement speech signals for each sub-tree determined to be inaccurate, and provide output based on corresponding text string sections of at least one sub-tree determined to be accurate.

The invention relates to a method for analysing mass spectral data obtained from a sample in GC×GC (2-dimensional) mass spectrometry, comprising: (a) comparing mass spectral data of an analyte with mass spectral data of candidate compounds of known structure in a data library; (b) identifying a plurality of candidate compounds from the library based on similarities of mass spectral data; (c) predicting, for each candidate compound, a value of at least one analytical property using a quantitative model based on a plurality of molecular descriptors; and (d) calculating a match score for each candidate compound based on the value predicted in step (c) and a measured value of the analytical property for the analyte.

There are provided methods and devices for improving a shopping experience of a user, including methods and devices for creating, updating, and maintaining a list such as a shopping list and methods and devices for automatically identifying a suitable substitute to a user selected product.

The invention discloses a hybrid position-posture tracking method based on inertia and optical tracking systems. Through the combined application of an inertia tracking system and an optical tracking system, the purpose of tracking the accurate position and posture of a target when a landmark is partially occluded can be achieved; the data of an inertial sensor is corrected by using optical tracking results, so that biased errors are eliminated, and the precision of an inertial tracking module in the process of posture estimation is improved; and when a kalman filter and the inertial tracking system are utilized for carrying out posture tracking, by establishing an acceleration error model and introducing a suddenly-changed-speed measurement noise covariance into a measurement noise covariance matrix, when the system has a linear motion, in a measurement updating equation of a state equation, the confidence level of a measured value is reduced, so that an estimated value in a state vector has a large confidence level, thereby reducing the influence of an accelerometer measured value on the state vector caused by the influence of a suddenly-changed linear acceleration.

A character based system and method for correcting low confidence characters from an OCR system facilitates operator review, editing and correction of character and field level data generated by an OCR system without the need for an application that is installed at the operator workstation. The system creates a data structure of OCR information and provides that information to an operator through an HTML interface that is rendered using HTML and JavaScript. The data structure includes an OCR confidence level for each character and / or field and the operator is prompted to review only those characters / fields that meet a predetermined threshold for the confidence level. The operator can use an input key (e.g., TAB or ENTER) to navigate to each character / field with a low confidence level and thereby correct or validate each low confidence character / field as appropriate.

A method and system for rating programs, the method including, receiving a sample of viewing logs from a plurality of set top boxes (STBs), determining, from the sample of viewing logs, at least groups of viewers sharing similar interests, and groups of programs sharing similar audience, computing time dynamics of rating distribution for the groups of viewers and the groups of programs, and incorporating at least one of the following into broadcast metadata the time dynamics of rating distributions of the groups of viewers for each of the groups of programs, rating distributions of the groups of viewers for each of the groups of programs marginalized by time, rating distributions of the groups of viewers for each program marginalized by time, relative sizes of each group of viewers, and a mapping of each program to groups of programs, thereby making the broadcast metadata available to the plurality of STBs for use in computing ratings. Related systems and methods are described.

A method and system for rating programs, the method including, receiving a sample of viewing logs from a plurality of set top boxes (STBs), determining, from the sample of viewing logs, at least groups of viewers sharing similar interests, and groups of programs sharing similar audience, computing time dynamics of rating distribution for the groups of viewers and the groups of programs, and incorporating at least one of the following into broadcast metadata: the time dynamics of rating distributions of the groups of viewers for each of the groups of programs, rating distributions of the groups of viewers for each of the groups of programs marginalized by time, rating distributions of the groups of viewers for each program marginalized by time, relative sizes of each group of viewers, and a mapping of each program to groups of programs, thereby making the broadcast metadata available to the plurality of STBs for use in computing ratings. Related systems and methods are described.

The invention discloses a Faster RCNN target detection method based on refractory sample mining. The method comprises the steps of 1, image target detection based on deep learning; 2, on the basis ofan online sample mining method, adjusting and setting adopted key parameters as follows: step 3, negative difficult-to-distinguish sample mining: on the basis of a difficult-to-distinguish sample, adjusting a mini-formed by an RPN in training; wherein the ratio of the positive sample to the negative sample of the bch is 1: 3; and step 4, removing redundant frames to avoid loss of multiple calculations. An improved non-maximum suppression algorithm is adopted to reasonably remove redundancy from a suggestion box generated by an RPN layer network. Under the condition that samples are not expanded. The definition of negative samples is broadened, and more samples difficult to train are mined online through the samples. The ratio of positive samples to negative samples is set, and rare sampleswith the maximum loss and difficult training are reasonably and simply calculated; and the loss of classification and frame regression is balanced, so that the continuous reduction of training loss can be met.

This invention is a method applicable to an image processing device, which includes the steps of providing a preprocess module for extracting a high-frequency portion of an image inputted into the device, extracting a gradient of the image and decomposing the image into plane and edge regions according to a predetermined fixed threshold, and providing a composite up-scaling module for executing the magnification processes on the image and the high-frequency portion thereof respectively, wherein the magnification process of plane regions of the image and the high-frequency portion is based on a simple interpolation while the edge regions of the image and the high-frequency portion is based on both a smart interpolation and the simple interpolation. The magnification results of the image and the high-frequency portion are then processed by a fusion process, so as to output an image having sharp but not blocky edges, rich details and strong contrast.

A system and related methods are disclosed for detecting the occurrence of an intrusion attack. A network device, such a probe, monitors traffic on a first network and converts the traffic to a format that is suitable for transmission on a second network. The converted traffic is forwarded to an intrusion detection system for further processing. Prior to transmission, the converted data may be filtered to remove data that is not useful in detecting an intrusion attack.

A method of generating a plurality of depth maps for a plurality of images comprises receiving a first image, obtaining information relating to the shot defined by the first image, generating a depth map for the first image according to a first schema, receiving a second image, obtaining information relating to the shot defined by the second image, detecting a change in the obtained information between the first and second image, and generating a depth map for the second image according to a second schema, the second schema having a complexity different from that of the first schema. The method can comprise accessing first and second depth models. In one embodiment, the first schema comprises the first depth model, and the second schema comprises the second model, and in a second embodiment the first schema comprises the first depth model, and the second schema comprises a combination of the first and the second depth models.

The invention provides a vehicle information matching method and system, and a monitoring device. The method comprises the steps: obtaining first information of electronic license plates of one or more vehicles and second information of physical license plates in a certain time period; judging whether there is second information accordance with certain first information or not; determining that the vehicle information matching of a vehicle corresponding to the first information is successful if there is the second information accordance with certain first information, or else, carrying out fuzzy matching of the first information with the unsuccessfully matched second information in the vehicle information. In other words, the method and system still can correct the second information in a fuzzy matching mode under the conditions that an external environment is severe (strong light, misty rain, or poor light), or a physical license plate cannot be recognized because the physical license plate is polluted or shaded (inaccurate second information of the physical license plate) and there is wrong recognition, and improve the accuracy of vehicle information recognition.

An image determination apparatus includes a frequency band signal detecting unit, an average value calculating unit, a relative value calculating unit, and an image determination unit. The frequency band signal detecting unit detects signals belonging to a plurality of frequency bands, from an image signal. The average value calculating calculates average values for respective ones of the signals belonging to the plurality of frequency bands detected by the frequency band signal detecting unit. Each of the average values is a characteristic value equivalent to an amplitude. The relative value calculating unit calculates a relative value of one of the average values to another of the average values, among the average values calculated for respective ones of the plurality of frequency bands. The image determination unit for determining an image based on the relative value calculated by the relative value calculating unit. The image determined is subject to image-processing.

A method for identifying words in a textual image undergoing optical character recognition includes receiving a bitmap of an input image which includes textual lines that have been segmented by a plurality of chop lines. The chop lines are each associated with a confidence level reflecting a degree to which the respective chop line properly segments the textual line into individual characters. One or more words are identified in one of the textual lines based at least in part on the textual lines and a first subset of the plurality of chop lines which have a chop line confidence level above a first threshold value. If the first word is not associated with a sufficiently high word confidence level, at least a second word in the textual line is identified based at least in part on a second subset of the plurality of chop lines which have a confidence level above a second threshold value lower than the first threshold value.

Various embodiments of a system for detecting airborne hazards in a gaseous environment are disclosed. An embodiment of the system includes a plurality of sets of one or more detectors, where the detectors of each set measure different types of properties of gases, such as the color emitted when the gases are exposed to certain reagents or the ionization potential of the gases. Environmental gases enter the system and are split into a plurality of air flow streams, with each stream directed to pass through one of the sets of one or more detectors. Data from all the sets of one or more detectors is sent to a computer programmed to integrate this data in order to determine information related to the identity (e.g., the specific identity or a classification) of chemical compounds (e.g., hazardous compounds) present in the gases.

A small target detection method based on SSD is provided by the invention. The feature information of a small target is enriched through multi-scale feature fusion, learning of high-resolution features with a large receptive field, introduction of an attention mechanism and other feature enhancement modes, the feature quality of the small target is improved, and the A nchor is learned through theconfidence of Anchor-object matching strategy joint classification and positioning, and the problem that no interaction exists between classification and positioning and between classification and positioning due to independent supervised network learning of each allocated Anchor is solved, so the detection precision and the positioning precision of the small target are improved; and in addition,the diversity of data set samples can effectively improve the precision and the generalization ability of the model for small target detection, so in the aspect of data sample collection, various influence factors such as different illumination, different postures and shielding degrees are referred.

A system and method for detecting and resolving mis-transcriptions in a transcript generated by an automatic speech recognition system when transcribing spoken words. The system and method receive a machine language generated transcript of a speech signal by at least one of a first machine learning system and a second machine learning system, and analyze the machine language generated transcript to find a region of low confidence indicative of a mis-transcription and predict an improvement to the region of low confidence indicative of the mis-transcription. The system and method select a replacement word for the mis-transcription based on the predicted improvement to the region of low confidence and replace the mis-transcription by the replacement word to generate a corrected transcript.

An approach for two stage log normalization is provided. The approach retrieves a message format and a plurality of parameters from one or more log files. The approach determines a classification for one or more first sequence files, wherein the one or more first sequence files includes the message format from the one or more log files. The approach determines a classification of error for the one or more first sequence files. The approach determines whether there is a high confidence in the classification of error for the one or more first sequence files. The approach determines whether there is an improvement in confidence in the classification of error from one or more second sequence files, wherein the one or more second sequence files includes the message format and the plurality of parameters from the one or more log files.

Method of comparing a test fingerprint with a stored reference fingerprint. The stored reference fingerprint and an image strip sequence of the test fingerprint are produced, wherein image strips of the image strip sequence represent different areas of the test fingerprint. Individual distance values between each image strip of the test fingerprint and at least one section of the reference fingerprint using a distance function are determined. An overall distance value from the individual distance values are determined, the overall distance value being a measure of similarity between the test fingerprint and the reference fingerprint.

The invention provides a vehicle identification method and a vehicle identification device. The method includes the following steps that: whether first information contained in the electronic license plate of a vehicle is matched with second information contained in the physical license plate of the same vehicle is judged, wherein the first information and the second information are acquired by a certain inspection gate, if the first information is matched with the second information, the vehicle is judged as a legal vehicle; if the first information is not matched with the second information, first information contained in the electronic license plate and second information contained in the physical license plate of the same vehicle acquired by a plurality of inspection gates which are adjacent to the inspection gate are obtained according to the driving path of the vehicle; and whether the first information and the second information which are acquired by each inspection gate are matched with each other is judged, if the first information is not matched with the second information, the vehicle is finally judged as an illegal vehicle. According to the method, when the second information cannot be clearly identified due to poor external environment, and the first information cannot be matched with the second information, the above judgment result can be verified through judging whether the first information and the second information of the same vehicle which are acquired by the plurality of inspection gates which are adjacent to the inspection gate are matched each other, and therefore, misjudgment caused by unclear identification of the physical license plate in a vehicle recognition process can be avoided.

In an inspection method and apparatus, when thermal image data with respect to cap sealing of a particular container is analyzed and a low confidence score is obtained, products are classified into a user check product by a rejection-process of separating the products, without determining the products to be good products or not good products.

The invention belongs to a data information processing technology, and discloses a deep learning adversarial sample generation method and system based on a second-order method, and the method comprises the steps: carrying out the secondary Taylor expansion of a neural network function in a tiny neighborhood of an input sample X, i.e., Lp (p belongs to [2, 0, infinity]) norm constraint, and replacing a nonlinear part of a neural network; and constructing a dual function through a Lagrange multiplier method to calculate an extreme value to solve the optimal disturbance delta, so that the confidence coefficient of the confrontation sample X '= X + delta which is judged as a correct class is reduced to the minimum, or the confidence coefficient of the confrontation sample X' which is judged asa target class is increased to the maximum. According to the method, the operation of reducing the confidence coefficient of the correct output class is adopted for the target-free attack; and for the target attack, the operation of improving the confidence of the target class is adopted. The method provided by the invention can avoid falling into a local extreme value, generates a high-quality adversarial sample at extremely low cost, is applied to adversarial training of the deep neural network, and can effectively improve the defense effect.