144results about How to "Reduce rejection rate" patented technology

The invention provides a speech recognition model establishing method based on bottleneck characteristics and a multi-scale and multi-headed attention mechanism, and belongs to the field of model establishing methods. A traditional attention model has the problems of poor recognition performance and simplex attention scale. According to the speech recognition model establishing method based on thebottleneck characteristics and the multi-scale and multi-headed attention mechanism, the bottleneck characteristics are extracted through a deep belief network to serve as a front end, the robustnessof a model can be improved, a multi-scale and multi-headed attention model constituted by convolution kernels of different scales is adopted as a rear end, model establishing is conducted on speech elements at the levels of phoneme, syllable, word and the like, and recurrent neural network hidden layer state sequences and output sequences are calculated one by one; and elements of the positions where the output sequences are located are calculated through decoding networks corresponding to attention networks of all heads, and finally all the output sequences are integrated into a new output sequence. The recognition effect of a speech recognition system can be improved.

This invention is a voice identification lock module and its working method. The module is composed by a keyboard, voice collecting module, voice processing module, switch status output module, voice notice module and RS232 communication port. The lock key is based on identification of individual voice character. Error rate is less than 1%.

A voice authentication system includes: a standard template storage part 17 in which a standard template that is generated from a registered voice of an authorized user and featured with a voice characteristic of the registered voice is stored preliminarily in a state of being associated with a personal ID of the authorized user; an identifier input part 15 that allows a user who intends to be authenticated to input a personal ID; a voice input part 11 that allows the user to input a voice; a standard template/registered voice selection part 16 that selects a standard template and a registered voice corresponding to the inputted identifier; a determination part 14 that refers to the selected standard template and determines whether or not the inputted voice is a voice of the authorized user him/herself and whether or not presentation-use information is to be outputted by referring to a predetermined determination reference; a presentation-use information extraction part 19 that extracts information regarding the registered voice of the authorized user corresponding to the inputted identifier; and a presentation-use information output part 18 that presents the presentation-use information to the user in the case where it is determined by the determination part that the presentation-use information is to be outputted to the user.

The invention discloses method for strengthening X2 interface resource state signaling, mainly solving the problem that resource state signaling of X2 interface interaction cannot indicate residual resource state information of a district in an LTE-Advanced system. The invention adopts the following strengthening processes: firstly, a source base station sends a resource state request message to an adjacent base station through an X2 interface for requesting the adjacent base station to periodically report a residual PRB resource state, and the adjacent base station feeds back a response message according to measuring capability per se; secondly, the adjacent base station periodically reports the residual PRB resource state to the source base station; and thirdly, the source base station sends the resource state request message and sets register request information to be 'stopping', and the adjacent base station feeds back the response message and ends work of report and measurement after receiving the message. The invention can realize interaction of the residual resource information of the district with low complexity without increasing an additional X2 interface application layer protocol progress. The residual resource information is a quantitative value, is accurate and reliable, and can be used for a district selection process for switching among honeycomb districts by skipping a district.

This invention discloses a method for laminating transmission nets, selecting joint routes and jointly optimizing multi-layer resource including the following steps: 1, expressing layering to the network, 2, jointly selecting routes according to the parameters of service request bandwidth and grade, link bandwidth usability, adjacent inter layer link transmission rate and hops, 3, calling the MRIS method based on the result of jointly selecting routes, 4, periodically calling the MRJS method to optimize the resource configuration, 5, carrying out the algorithm operation.

A method for automatically changing a slag line of a tundish submersed nozzle for an ultra-thick plate blank casting machine belongs to the technical field of plate blank continuous casting. The process steps are: installing a position sensor on a hydraulic cylinder of a tundish trolley and obtaining average values of actual level of the tundish trolley in a steel pouring process through numerical values reflected by the position sensor; installing an automatic liquid lever measuring meter on a crystallizer, and obtaining actual crystallizer level in the steel pouring process through a crystallizer liquid level automatic measuring system; installing the submersed nozzle on the tundish trolley, and measuring the distance between the bottom of the tundish trolley and the top of the submersed nozzle; lowering the tundish trolley to a lowest position, and measuring the distance between the bottom of the tundish trolley and the top of a crystallizer copper plate; inputting a formula for calculating actual immersion depth of the submersed nozzle in the steel pouring process into a software program of the casting machine, setting time intervals and changing amount of each time of height changing of the tundish trolley in an automatic system of the casting machine and achieving the operation of automatic slag line changing of the submersed nozzle. The method has the advantages of prolonging service life of the tundish submersed nozzle of the casting machine.

The invention discloses a self-service claim settlement method, which comprises the steps of obtaining the identity and report information of an insured person according to a report request upon receiving the input report request; obtaining all policies of the insured person according to the identity of the insured person; according to the report information, screening out the claimable policy ofthe report in the all the policies; and generating a corresponding claim settlement case according to the claimable policy and the report information. The invention also discloses a server and a computer readable storage medium. According to the invention, the automatic screening of the policy for applying for claim settlement is realized, the proportion of claims rejected is reduced, and the aging time of claim settlement is improved.

The invention discloses an imaging distortion correction method and device, a computer device and a storage medium. The method comprises the following steps of calculating a theoretical image height when the image is zero distortion according to the field angle of view of a wide angle lens, the distortion rate of the wide angle lens and the actual image height after the image distortion formed bythe wide angle lens; generating a first correction function for correcting the distortion of the image formed by the wide angle lens according to the theoretical image height and the field angle of view; compensating a first correction function for obtaining a second correction function for correcting distortion of an image formed by an optical module of a wide angle lens, wherein the wide-angle lens optical module comprises a wide-angle lens; acquiring the image formed by the wide-angle lens optical module, and correcting the image formed by the wide-angle lens optical module by a second correction function. The invention can accurately correct the image formed by the optical module of the wide-angle lens, reduces the rejection rate and the recognition rate of the image when the image isrecognized, and improves the user experience.

The embodiment of the invention provides an identity registration and identity authentication method, device and system. The method comprises the steps that according to a login password and login biological characteristics, a terminal is used for extracting recovery biological characteristics from a biological characteristic safe box by the adoption of a biological characteristic decoding algorithm; then, the terminal is used for obtaining registration verification information, and generating login verification information by the adoption of a registration verification algorithm according to the recovery biological characteristics and the login password, and if the registration verification algorithm is identical to the login verification information, the terminal determines that the login password and the login biological characteristics are effective. In the method, when a user is legal, the recovery biological characteristics extracted from the biological characteristic safe box according to the login biological characteristics are completely identical to registration biological characteristics, and therefore the authentication rejection rate of biological characteristic authentication can be decreased, and the recognition capability of the biological characteristics is enhanced.

The invention discloses an efficient routing method under a distributed domain-division network, which comprises the following steps that: a source node sends a service request packet to a cluster head node of a region, and the cluster head node judges whether the source node and a target node are positioned in the same region or not; if the source node and the target node are in the same area, the cluster head node calls area resource pool information and returns one or more routing paths meeting the QoS (Quality of Service) requirement of the service based on a routing decision algorithm ofreinforcement learning, and end-to-end routing establishment of the service is completed; if the source node and the target node are in different regions, the cluster head node calls global resource pool information, judges the region through which the service needs to pass based on a routing decision algorithm of reinforcement learning, and splits the service request by adopting a cross-region service request splitting method; and the split service requests are sent to the cluster head nodes of the passing areas, then area resource pool information is called, and finally end-to-end routing decision is completed by multi-area cooperation based on a routing decision algorithm of reinforcement learning to obtain a routing path from a source node to a target node.

The invention discloses a fingerprint template update method, a fingerprint template update device and terminal equipment. The method comprises steps that a fingerprint image inputted by a user is matched with fingerprint templates inputted in a fingerprint database, and the inputted information corresponding to the fingerprint templates is marked; if a matching degree of the fingerprint image and the fingerprint templates is greater than a preset threshold, whether the quantity of the inputted fingerprint templates reaches a preset threshold value is detected; if the quantity of the inputted fingerprint templates reaches the preset threshold value, to-be-replaced fingerprint templates are deleted from the fingerprint database according to the inputted information corresponding to the fingerprint templates; the fingerprint image is taken as a new fingerprint template to input to the fingerprint database, and the inputted information corresponding to the fingerprint image is further marked. Through the method, the fingerprint templates can be timely and sequentially updated, a false rejection rate is reduced, and validity of a fingerprint learning function and fingerprint identification accuracy are improved.

The invention discloses a processing process for threshing, re-drying and classifying tobacco stems, which comprises the following steps: firstly performing leaf-stem separation; preliminarily screening stem turning heads and broken stems; then screening the tobacco stems into the three types of tobacco stems with different thicknesses of 1.5-2.0 mm, 2.0-2.5 mm and 2.5-3.0 mm; pre-pressing the three types of tobacco stems, wherein gaps among the pressed stems are 0 .8 mm, 1.0 mm and 1.2 mm respectively; mixing the three types of pre-pressed tobacco stems and re-drying; finally packaging and storing. According to the processing process disclosed by the invention, the tobacco stem re-drying efficiency can be improved, the consistence of the batch water loss rate and the moisture content after re-drying is improved, and the tobacco stem aging progress is accelerated; the tissue destruction difference degree of the tobacco stems with different thicknesses in the cut stem making process is reduced, the stem storage time is shortened, the production efficiency of the cut stem making procedure is improved, the cut stem moisture content fluctuation, the rejection rate and the energy consumption are reduced, and the uniformity of cut stems, the moisture content and the stability of filling values are improved.

An underground coal mine deep-hole directional blasting cartridge comprises a cartridge body (1), an upper connector (2) and a lower connector (3). One side or opposing sides of the cartridge body (1) are provided with cuts (4), and each cut is covered with a layer of metal sheet (5) which is provided with a V-groove by concaving; the bottom of every V-groove points to the center of circle along the peripheral radial direction of the cartridge body and is arranged at a 1/2 radius part. The underground coal mine deep-hole directional blasting cartridge allows for more ideal directional blasting.

The invention relates to a self-adaptive fingerprint identification system which comprises a fingerprint acquisition unit. The fingerprint acquisition unit is provided with a fingerprint analysis module for carrying out focusing processing on an acquired fingerprint. According to the self-adaptive fingerprint identification system provided by the invention, the fingerprint analysis module is arranged for carrying out focusing processing on the acquired fingerprint, so that ideal fingerprint imaging can be ensured to be obtained, the influence of the environment on the fingerprint identification can be reduced, the accuracy of the fingerprint identification is improved and the false reject rate of inputting the fingerprint is reduced.