638 results about "Data Noise" patented technology

An implantable medical device (IMD) capable of monitoring physiologic data, distinguishing relatively noisy and noise free physiologic data, and recording noisy and relatively noise free segments of physiologic data in separate memory registers of a limited memory for retrieval and analysis at a later time. Preferably the physiologic data comprises the sampled EGM of the heart detected from sense electrode pairs that are implanted in the patient at sites where extraneous electrical noise, e.g., electromyographic signals, are also capable of being detected. The sense electrode pairs can constitute one or both sense electrodes located on or adjacent to the atrial and / or ventricular heart chambers and coupled to the IMD by a lead body or sense electrode pairs that are located remotely from the heart, e.g. at a subcutaneous implantation site of the IMD. A plurality of noisy EGM episode data registers store a corresponding plurality of noisy EGM episode data sets on a FIFO basis and another plurality of noise free EGM episode data registers to store a corresponding plurality of relatively noise free EGM episode data sets on a FIFO basis. Any form of discrimination of noisy data from relatively noise free data can be employed at the time of recording, but because the stored EGM episode data sets are subsequently viewed and analyzed by a physician, discrimination with absolute certainty is not required, and the physician can alter the detection criteria to fine tune it.

A human assisted method of debugging training data used to train a machine learning classifier is provided. The method includes obtaining a classifier training data set. The training data set is then debugged using an integrated debugging tool configured to implement a debugging loop to obtain a debugged data set. The debugging tool can be configured to perform an estimation and simplification step to reduce data noise in the training data set prior to further analysis. The debugging tool also runs a panel of prediction-centric diagnostic metrics on the training data set, and provides the user prediction based listings of the results of the panel of prediction-centric diagnostic metrics.

Triggers and noise should be available as information in recorded electrograms in memories of implantable medical devices. Particularly where the recording of electrogram data is done in the far field, there will be considerable noise and the interpretation of ECG's reproduced from such recorded data will benefit from the storing of information regarding contemporaneous noise. By storing contemporaneous trigger data and noise data directly in the ECG data, recordings of the ECG data become more useful for physician use when played back through an external display system with minimal loss of ECG data, since out of range values are employed for the noise and trigger information and this non-ECG data is limited in size to no longer than individual point values of the ECG signal.

Systems and methods for detection and geolocation of multiple emitters that are emitting RF signal energy on a common frequency, and that may be implemented to separate, geolocate, and / or determine the number of emitters (e.g., radio users) emitting on a common RF frequency. Real-time signal qualification processing may be employed to continuously monitor and collect incoming receiver tuner data for signal activity and ignore irrelevant noise data. Each set of data blocks from an emitter transmission signal may be defined as an emission cluster, and a set of time difference of arrival (TDOA) / frequency difference of arrival (FDOA) pairs may be computed for each emission cluster with each TDOA / FDOA pair yielding a geolocation result. A statistical qualification method may be used to produce a final geolocation answer from each set of emission cluster geolocation results, and a geolocation error ellipse computed for the final geolocation answer of each emission cluster. The final geolocation answer of each emission cluster may be run through a correlation and fusion algorithm.

Noise reduction in a one- or two-dimensional capacitive sensor array is achieved by rejecting and re-acquiring noisy signals. The sensor array is formed of crossed X and Y lines for drive and sense functions respectively, each of the X lines being driven in turn to acquire a full frame of data from the sensor array. A controller actuates the X lines in turn and, for each X line, charge is transferred to charge measurement capacitors connected to respective ones of the Y lines. The controller measures a signal value from a first one of the measurement capacitors, and then tests if that Y signal value has a magnitude lying within an acceptable range. If not, the measurement capacitors are all reset without their signal values being measured, and the controller re-drives the same X line to initiate another charge transfer into the measurement capacitors for that X line. It is then attempted once again to acquire Y signal values for that X line. Noisy signal values are thus rejected, and the sensor re-acquires substitute signal values. This is distinct from the approach of acquiring data and then reprocessing it to remove or suppress noisy data as in the prior art.

In one example embodiment, a method is illustrated as including retrieving item data from a plurality of listings, the item data filtered from noise data, constructing at least one base cluster having at least one document with common item data stored in a suffix ordering, compacting the at least one base cluster to create a compacted cluster representation having a reduced duplicate suffix ordering amongst the clusters, and merging the compact cluster representation to generate a merged cluster, the merging based upon a first overlap value applied to the at least one document with common item data.

A method of image processing using kernel regression is provided. An image gradient is estimated from original data that is analyzed for local structures by computing a scaling parameter, a rotation parameter and an elongation parameter using singular value decomposition on local gradients of the estimated gradients locally to provide steering matrices. A steering kernel regression having steering matrices is applied to the original data to provide a reconstructed image and new image gradients. The new gradients are analyzed using singular value decomposition to provide new steering matrices. The steering kernel regression with the new steering matrices is applied to the noisy data to provide a new reconstructed image and further new gradients. The last two steps are repeated up to ten iterations to denoise the original noisy data and improve the local image structure.

The invention relates to a method for acquiring dynamic traffic information based on a middleware, which overcomes the shortcomings of data loss, data noise, and particularly the multi-source isomerism of data, reduces redundant data, ensures the accuracy of the data and improves the accuracy and the reliability of the data. The method comprises the following steps of: 1) transmitting the traffic information by adopting a serial interface communication mode and / or a network communication mode; 2) customizing an information acquisition port which can be matched with different inspection devices of the traffic information by using an interface definition language IDL in the CORBA middleware technique, identifying and normalizing the data from different inspection devices, thus acquiring the real-time dynamic traffic information, such as road traffic flow, vehicle velocity, road occupancy rate and the like; 3) performing preprocessing on the all data acquired by the method, and performing map matching on a floating automobile by using a road matching algorithm based on a network topology relationship; and 4) fusing and saving the multi-source isomerous real-time dynamic traffic data which is preprocessed by the method in a database by using an immune clustering neural network.

The invention discloses an electrocardiogram diagnosis method and system based on a deep convolution neural network. The diagnosis method comprises the steps that a collected original electrocardiogram of the human body is preprocessed to obtain a plurality of heartbeat fragments with a preset length; the heartbeat fragments are input into a trained deep convolution neural network model, and a prediction value of each category is obtained; the category of a single heartbeat corresponding to the maximum prediction value and the time interval of the heartbeat and a previous heartbeat are recorded, and a time series is formed; on the basis of a preset mapping relation between a clinical diagnosis criterion and the time series, the time series is matched with the diagnosis criterion, and a corresponding diagnosis result is obtained. Any feature in the electrocardiogram is not required to be extracted, so that valid information loss, noise data introduction and other problems caused by manual screening and processing are avoided to the largest extent, and the influence of manual characteristic extraction on the diagnosis accuracy is reduced.

An object of the present invention is to facilitate dealing with noisy speech with varying SNR and save calculation costs by generating a speech model with a single-tree-structure and using the model for speech recognition.Every piece of noise data stored in a noise database is used under every SNR condition to calculate the distance between all noise models with the SNR conditions and the noise-added speech is clustered. Based on the result of the clustering, a single-tree-structure model space into which the noise and SNR are integrated is generated (steps S1 to S5). At a noise extraction step (step S6), inputted noisy speech to be recognized is analyzed to extract a feature parameter string and the likelihoods of HMMs are compared one another to select an optimum model from the tree-structure noisy speech model space (step S7). Linear transformation is applied to the selected noisy speech model space so that the likelihood is maximized (step S8).

The present invention provides a method for analyzing emotional polarity of heterogeneous migration images based on multimodal depth potential correlation. The method comprises the following steps: 1)constructing an initial emotional image dataset, and taking the emotional polarity corresponding to the emotional words as the emotional polarity tag; 2) removing the noise data in the initial emotion image data set, and removing the noise by using the method of emotional consistency and the probabilistic sampling model based on the multimodal deep convolution neural network; 3) constructing theheterogeneous migration model based on the multimodal depth potential correlation, and then training the source domain text and the target domain image; 4) constructing the multimodal embedded space,embedding semantic information of the source domain text into the target domain image; and 5) training the image emotional polarity classifier for the image emotional polarity analysis. According to the method provided by the present invention, the obtained data is large in scale, the labor cost is low, the data noise is small, the prediction accuracy is high, the model is strong in interpretability and has strong classification capability, and a better image emotional polarity analysis effect can be reached.

Methods, systems and computer program products for simplifying complex data stream problems involving feature extraction from noisy data. Exemplary embodiments include a method for processing a data stream, including applying multiple operators to the data stream, wherein an operation by each of the multiple operators includes retrieving the next chunk for each of set of input parameters, performing digital processing operations on a respective next chunk, producing sets of output parameters and adding data to one or more internal data stores, each internal data store acting as a data stream source.

Methods and systems for recommending a service based on an electronic social media trust model. A user trust network and a service trust network can be constructed and the two separate trust networks can be combined to form a combined trust network. The combined trust network includes an explicit trust and an implicit trust in order to improve the recommendation coverage and consider a latent service rating without suffering noisy data. A trust-oriented random walk model can be conducted on a user node with respect to the combined trust network based on a user search intent and navigation behavior in order to select and recommend a service candidate. A service rating can then be predicted by considering the user ratings with respect to a target service, a propagated trust and an inferred service rating.

An automatic peak selection method for multidimensional data that selects peaks from very noisy data such as two-dimensional liquid chromatography-mass spectrometry (LC-MS) data is described. Such data are characterized by non-normally distributed noise that varies in different dimensions. The method computes local noise thresholds for each one-dimensional component of the data. Each point has a local noise threshold applied to it for each dimension of the data set, and a point is selected as a candidate peak only if its value exceeds all of the applied local noise thresholds. Contiguous candidate peaks are clustered into actual peaks. The method is preferably implemented as part of a high-throughput platform for analyzing complex biological mixtures.

Methods and apparatus for estimating and correcting carrier frequency offsets in a bust multi-tone receiver are described. Course and fine carrier frequency estimates are generated from the signal's preamble. Decision directed carrier frequency offset estimates are then generated from the signal field and data fields of the multi-tone signal. Frequency error estimates are generated for each tone of the signal and combined using a weighted average to generate the frequency error estimate used to perform the correction operation. Error estimates corresponding to noisy data tones are weighted less then estimates corresponding to less noisy data tones. In cases of low SNR frequency error estimates corresponding to pilots are weighted by an extra amount as compared to error estimates corresponding to tones used to transmit data symbols. During times of high SNR error estimates corresponding to pilot tones are weighted in the same manner as error estimates corresponding to data tones.

The invention provides an industrial soft gateway based on multiple access and edge computing and an implementation method. The industrial soft gateway comprises a configuration interaction module, adata collection module, a data edge computing module and a data sending control module, wherein the configuration interaction module comprises a connection configuration module and a data standardization module; the data collection module is used for a multiple access collection method; the data edge computing module is used for carrying out real-time computing on the data collected by the data collection module; and the data sending control module is used for carrying out caching and task scheduling and allocation of external forwarding on all data to be sent. According to the industrial softgateway based on multiple access and edge computing and the implementation method, a visual interface is arranged to provide connection configuration and standardized operation for various communication protocols and databases, and meanwhile, a graphical monitoring interface of a data collection state, an edge computing result and a forwarding state is provided. Data standardization can be realized, the data noise can be eliminated and the data features can be extracted through the edge computing module, the network transmission data volume of a data cloud platform is reduced, and the data transmission efficiency is improved.

The invention relates to the technical field of computer vision, in particular to an instance-level image search method based on multiple layers of feature representations. According to the method, feature representations, on different layers, of images are learnt through a deep convolution neural network model for instance-level image search so that different images of the same article can be effectively found. A coding learning process is introduced on the basis of a traditional network model, and features from multiple convolution layers are automatically encoded, so that the extracted features are more robust, and influences of background and noise data on the features are reduced. Meanwhile, the features integrate local information and class information. The invention further discloses a loss function based on multiple tasks. By optimizing the function, the learnt features have good generalization performance, and the learnt features can be well used for distinguishing inter-class images from intra-class images of different things.

A method and apparatus for ensuring a level of privacy for answering a convolution query on data stored in a database is provided. The method and apparatus includes the activities of determining (402) the level of privacy associated with at least a portion of the data stored in the database and receiving (404) query data, from a querier, for use in performing a convolution over the data stored in the database. The database is searched (406) for data related to the received query data and the data that corresponds to the received query data is retrieved (408) from the database. An amount of noise based on the determined privacy level is generated (410) and added (412) to the retrieved data to create noisy data which is then communicated (414) to the querier.

Example embodiments of systems and methods for network pattern matching provide the ability to match hidden networks from noisy data sources using probabilistic multi-attribute graph matching analysis. The algorithms may map roles and patterns to observed entities. The outcome is a set of plausible network models. The pattern-matching methodology of these systems and methods may enable the solution of three challenges associated with social network analysis, namely network size and complexity, uncertain and incomplete data, and dynamic network structure.

The invention provides a marketing analysis data market system. The marketing analysis data market system comprises a data access layer, a data extraction module, a data conversion module, a data cleaning module, a log and alarm sending module and a data downloading module, a data packet of the data access layer contains office data, external data and service data, and models of the system includes a data logic model and a data physics module. Firstly, necessity in designing the marketing data market is analyzed and ETL data processing including noise data processing, data uniformity and data quality and the like is analyzed by discussing a data integration method, and various data sources can be reorganized and processed by a data transition tool. In addition, the physics model in the data market is realized according to the physic list structure of the logic model. Finally, application prospect of the data market in the marketing analysis is expected.

The invention relates to the technical field of mobile internet, and discloses a method and a system for analyzing network behaviors of mobile terminals users. The method includes collecting access logs and signaling log data of the users to the mobile internet through a collection device, cleaning the collected data to form an effective data set, classifying and analyzing the data set through a matched mobile internet knowledge base, analyzing users' behaviors by loading a user behavior analysis algorithm to acquire analysis results of users' behaviors, and finally displaying the analysis results to the users. According to the collected network requests sent from mobile terminals and the signaling data, the users can be recognized accurately; by means of building the mobile internet knowledge base, users' behavior preferences can be found accurately; therefore, problems such as high data noises, small data sample and inaccurate prediction in the prior art are solved.

Noise-superimposed speech data is grouped according to acoustic similarity, and sufficient statistics are prepared using the speech data in each of the groups. A group acoustically similar to voice data of a user of the speech recognition is selected, and sufficient statistics acoustically similar to the user's voice data are selected from the sufficient statistics in the selected group. Using the selected sufficient statistics, an acoustic model is prepared.

Systems and methods are presented that enable logical reasoning even in the presence of noisy (inconsistent) data. The knowledge base is processed in order to make it consistent and is also compiled. This processing includes checking and correcting spelling, removing stopwords, performing, grouping words of similar and related meaning, and compacting the knowledge base. A robot can use the processed knowledge base to perform many different types of tasks, such as answering a query, determining a course of action that is designed to achieve a particular goal, and determining its own location.

A method for monitoring engine performance includes sampling exhaust gas temperature associated with a turbine engine over an interval of operational time of the turbine engine. The method further includes applying a first test to identify statistical outliers on the sampled exhaust gas temperature data and removing identified statistical outliers from the sampled exhaust gas temperature data. Subsequently, the method includes applying a second test to identify step changes in slope of the exhaust gas temperature data and dividing the interval of operational time into one or more segments based upon the identified step changes. Finally, the method includes determining a slope for each segment and combining the segments to obtain a rate of performance deterioration of the turbine engine.

The invention discloses a relation extraction method in combination with clause-level remote supervision and semi-supervised ensemble learning. The method is specifically implemented by the following steps of 1, aligning a relation triple in a knowledge base to a corpus library through remote supervision, and establishing a relation instance set; 2, removing noise data in the relation instance set by using syntactic analysis-based clause identification; 3, extracting morphological features of relation instances, converting the morphological features into distributed representation vectors, and establishing a feature data set; and 4, selecting all positive example data and a small part of negative example data in the feature data set to form a labeled data set, forming an unlabelled data set by the rest of negative example data after label removal, and training a relation classifier by using a semi-supervised ensemble learning algorithm. According to the method, the relation extraction is carried out in combination with the clause identification, the remote supervision and the semi-supervised ensemble learning; and the method has wide application prospects in the fields of automatic question-answering system establishment, massive information processing, knowledge base automatic establishment, search engines, specific text mining and the like.

In a method and system for analyzing and predicting bidding of electric power generation, bid data is collected for a selected region. The bid data is preprocessed to decode the identity of the bidding entities and / or to remove erroneous or noisy data. External information, such as fuel prices, weather data, and / or generator outage data, is also collected. A bid prediction model is generated for one or more of the bidding entities from the bid data and the external information. External information is subsequently received, and the bid prediction model is applied to that current external information to predict future bids for one or more of the bidding entities. The predicted future bids are then reported.

The invention discloses a method for establishing a mapping knowledge domain based on a book catalogue. The method comprises the steps that a catalogue page in a digitized book is extracted, the lengths of items in the catalogue are differentiated, and part-of-speech tagging is conducted on the long items through a natural language processing tool, so that part-of-speech arrays are obtained, and candidate nodes are extracted according to rules of conjunctions, punctuations and parts of speech; the long items and the short items are authenticated in the Baidu encyclopedia and the Hudong encyclopedia, a leader-member relation and parallel relations are formed through a catalogue structure and serve as a framework of the mapping knowledge domain, the strong and weak parallel relations are differentiated and serve as increments respectively, and the leader-member relation is supplemented with the strong and weak parallel relations; according to a noisy data excavating algorithm with suffixes serving as a base, nodes are selected from the items which do not pass the authentication of the encyclopedias and the mapping knowledge domain is supplemented with the selected nodes; finally, the weights of relations in the supplemented mapping knowledge domain are calculated and ranked, so that noise is removed through screening. Compared with an existing mapping knowledge domain, the mapping knowledge domain established through the method is richer in node, better in expandability and higher in accuracy.

The invention discloses a method for predicting industrial sewage inflow based on an ARIMA model. The method comprises the following steps: analyzing initial time series data to meet a requirement onARIMA model establishment; preprocessing abnormal data by eliminating, filling and the like; removing data noise through moving average filtering; by a unit root testing method, performing ADF testingon the stability of a time series; analyzing and verifying the non-randomness through an autocorrelation coefficient; preliminarily determinating an autoregressive and moving average order of the ARIMA (p, d, q) model, and then performing order determination on the model through combination with an AIC information criterion; optimizing model parameters by a least squares method; finally, testinga residual and evaluating a simulation result to determine a final prediction model. Acquired sewage inflow data are determined, and the obtained prediction model is used for predicting test data, andan output of the model is a prediction result of the sewage inflow. By the method, the model is succinct, the fitting effect of the prediction model is very good, and the precision is high.

The invention discloses a method for microblog hot event online detection based on emotion analysis. According to the method for microblog hot event online detection based on emotion analysis, emotion extraction is conducted on online microblog texts according to emotion analysis, the bursting state of emotions and emotion texts is detected according to the improved Kleinberg algorithm, hot topics are extracted according to an affinity propagation clustering algorithm in the bursting period, so that only a small part of texts with a large amount of effective information participate in event extraction, the data volume is reduced, meanwhile a large amount of data noise is filtered out, and therefore the purpose that online real-time detection of the microblog hot events is achieved.

A human assisted method of debugging training data used to train a machine learning classifier is provided. The method includes obtaining a classifier training data set. The training data set is then debugged using an integrated debugging tool configured to implement a debugging loop to obtain a debugged data set. The debugging tool can be configured to perform an estimation and simplification step to reduce data noise in the training data set prior to further analysis. The debugging tool also runs a panel of prediction-centric diagnostic metrics on the training data set, and provides the user prediction based listings of the results of the panel of prediction-centric diagnostic metrics.