562results about How to "Reduce false recognition rate" patented technology

The invention discloses a method for recognizing voice. The method comprises the following steps of: acquiring audio data; acquiring a Lattice result of the audio data, wherein the Lattice result comprises time point information, a plurality of pieces of candidate information and matching likelihood scoring information; acquiring confidence scoring information according to the plurality of piecesof candidate information and the matching likelihood scoring information; rearranging the plurality of pieces of candidate information by using a stronger voice model and providing the optimal recognition result; positioning a voicing position corresponding to the audio data and simultaneously displaying other candidate words; selecting or inputting a correct text to finish amendment and freezingthe amended text; and searching a related text training language model by using a search engine according to the amended text serving as a key word, interpolating to acquire an adaptive language model, and returning and newly recognizing the rest part of audio data by using the adaptive voice model. By using the technical scheme, the voice recognition rate can be improved, and the workload of manual checking can be reduced.

The invention relates to a multi-mode biometric authentication method and a multi-mode biometric authentication system. The system comprises an image acquisition module, a voice acquisition module, a main processing module and a storage module, wherein the main processing module further comprises a control unit electrically connected with the image acquisition module and the voice acquisition module, as well as a face recognition unit, a voice recognition and a lip shape and voice synchronous detection unit which are electrically connected with the control unit; the storage module is respectively connected with the face recognition unit and the voice recognition unit; the control unit is used for sending biological characteristic images or voice information received by the image acquisition module and the voice acquisition module to the face recognition unit, the voice recognition unit or the lip shape and voice synchronous detection unit according to the current recognition process of an identity authentication device so as to perform recognition, so that a recognition result is obtained. The method and the system have the advantage that by virtue of integrating three biological characteristics of face, lip shape and voice and performing pronunciation characteristic and lip shape characteristic synchronous detection, the system recognition error ratio is reduced.

The invention provides a speech recognition apparatus, a method and a piece of electronic equipment. The apparatus comprises an identifying unit used for identifying a speech to acquire a candidate keyword; a decoding unit decoding speeches including the speech of the identified candidate keyword on the basis of meaning information to generate a word grid corresponding to the speeches including the speech of the identified candidate keyword; a calculating unit calculating the confidence of the candidate keyword according to the word grid; and a judging unit judging whether to determine the candidate keyword as a keyword according to the confidence. By performing keyword identification in a way to refer to the meaning information, the apparatus solves the mistaken identification problem caused by similar pronunciations.

The invention provides a deep learning-based traffic sign automatic identifying and marking method, which is used in the technical field of environmental perception of intelligent vehicles. A semantic segmentation structure is adopted for locating and detecting traffic signs, and candidate regions with the traffic signs are obtained, wherein the semantic segmentation structure comprises two parts: an encoding network; and a decoding network and a pixel-based classification layer. Then, the traffic signs in the candidate regions are subjected to classified identification and locating through a convolutional neural network of a quick region. Based on the traffic sign automatic identifying and marking method, the invention furthermore correspondingly provides an effective map navigation information updating method. According to the method, the candidate regions with the traffic signs are located by using the semantic segmentation method, so that a new idea is provided, a training parameter quantity is reduced, the storage space is saved, and the computing time is shortened; and the method is high in identifying accuracy, and can perform relatively accurate traffic sign information updating on map navigation information, thereby better serving drivers conveniently.

The invention relates to the field of traffic image processing, in particular to a bayonet vehicle image identification method based on image features. The steps of bayonet vehicle image recognition method based on the image features includes that a template dada bank is established, the data bank stores template images which are shot with different style and different color vehicles, vehicle attribute data, vehicle body color data and vehicle scale invariant feature transform (SIFT) feature data are stored corresponding to each template image, wherein the SIFT feature data are obtained through data processing; vehicle recognition is conducted to to-be-identified images, color identification is conducted to the to-be-identified images, the template images which are fitted with the vehicle body colors are selected form the template data bank according to color identification results; the to-be-identified images are extracted the SIFT features and compared with the selected template images; the template images which are matched with the to-be-identified images are obtained; the vehicle attribute data are output, wherein the vehicle attribute data correspond to the matched template images. The bayonet vehicle image identification method based on the image features combines the vehicle color identification and the recognition based on a SIFT operator principle. Color information is added in the identification process. The defect that the SIFT operator discard the color information is overcome. Identification accuracy can be improved.

A method for identifying voice emotion based on support vector computer includes characteristic picking up-analyzing to collate characteristic parameter selection of global structure and sex, as well as to collate characteristic parameter selection of time sequence structure and to collate sex and vowel number; support vector computer training to carry out identification on five emotions of happy, angry, sad, fear and surprise.

The invention belongs to the technical field of the video processing, and particularly relates to a method for identifying content of a digital video, which mainly comprises a key frame extraction module, a video copy detection module and a video frame content identification module. The video copy detection module adopts a video copy detection method based on local ordering to ensure that a similar frame is rapidly obtained through a given possibility, and to improve the detection performance. The video frame content identification module adopts a pornographic image identification algorithm, the algorithm comprises four weak classifiers (skin area proportion classifier, face quantity classifier, invariant-moment classifier and skin area distribution classifier), the four weak classifiers are integrated to form one strong classifier, the result of the strong classifier is the final detection result, so the mistake recognition rate and nonrecognition rate can be effectively reduced. The method has wide application value for the internet digital video information monitoring.

The invention discloses an abnormal behavior detection method, which comprises the following steps of: extracting dynamic human skeleton joints in a video by using a neural network human skeleton extraction model to form a skeleton data set; obtaining a higher-level behavior feature map corresponding to the bone, namely surface behavior features, by an ST-GCN network; and inputting the behavior feature map into the abnormal behavior classifier model, and performing matching to identify the behavior type. The invention also discloses an abnormal behavior detection system which comprises a videomonitoring module and a network model integration module. According to the method, various human body behaviors and a large amount of human body skeleton data can be accurately and efficiently processed, and abnormal behaviors appearing in video monitoring can be automatically identified.

The invention discloses an anti-counterfeiting fingerprint identification device and an anti-counterfeiting fingerprint identification method. The device is provided with a sense control unit which comprises a CapSenseSigma-Delta (CSD) chip, wherein the CSD sense chip is connected with a conductive film. Fingerprints of living organisms are judged by recording an increment in the CSD sense chip and summating numeric codes converted by a modulator. The fingerprint identification method comprises the following steps in turn: 1) initializing; 2) circuit self-checking; 3) performing a corresponding program according to information input by a user; 4) judging whether to communicate; 5) acquiring image data; and 6) comparing registered users. The device of the invention has low cost, can outputa signal by touching lightly, does not need to adopt any distortion correction optical element, and can reduce manufacturing cost and realize miniaturization; the linear distortion relationship obtained by comparing an acquired fingerprint image with an actual fingerprint image can be corrected by software; and therefore, the device has the advantages of high quality of acquired optical imaging image, high fingerprint identification reliability, low error identification rate, good effect on anti-counterfeiting fingerprints, simple structure and low power consumption.

The patent discloses an intersection vehicle violation detection and identification method based on deep learning. The method comprises the following steps of 1, collecting intersection video detection images; 2, identifying traffic signs and vehicles; 3, tracking the vehicles; and 4, performing vehicle violation detection and identification. Based on automatic identification and tracking of the traffic signs and the vehicles in the intersection monitoring videos, the violation conditions of vehicles at the intersection can be automatically identified, and the system has few false detection and high efficiency.

The invention relates to a cascaded neural network-based face recognition method. The method comprises the following steps that: a) a training-used face images set is established; b) key points of each face image are detected; c) key point information is utilized to normalized the images into standard gray-value images; d) the standard face images are subjected to gray value statistics normalization; E) a depth neural network is trained, and input is a pair of images, and outputted 0 indicates that the two images are judged as a negative case pair, and outputted 1 indicates that the two images are a positive case pair; f) all image pairs are inputted into the neural network, and the proportion of negative case pairs which are wrongly judged as 1 is made to be lower than a given proportion; g); the step e) is repeatedly used; f), the total number of networks achieves a predetermined number, the method is terminated; and h) any pair of new face images is given, and an obtained cascaded neural network is utilized to judge whether the one pair of new face images is a negative case pair or a positive case pair. With the cascaded neural network-based face recognition method of the invention adopted, the recognition rate of face recognition can be effectively improved, and wrong recognition rate can be greatly reduced, and the reliability of a security monitoring system can be greatly improved.

The invention relates to a recognition method of a license plate number of a motor vehicle, and belongs to the technical field of license plate number recognition. The recognition method comprises carrying out median filtering and noise reduction to an input image containing a license plate, locating multiple license plate areas by utilizing a binarization processing method and a morphological processing method, carrying out rotation correction and edge removing processing to the license plate areas one by one, carrying out character cutting by utilizing a vertical projection method, carrying out further edge removing processing to obtained character blocks in a cutting mode, carrying out character recognition by utilizing an enhanced edition template matching method, and therefore achieving fast and accurate license plate recognition. The recognition method improves the precision of license plate location, and fast and precise location of the license plate is achieved. The recognition method can be used in a real-time license place recognition system, and for processing high-speed data flow, meanwhile is high in recognition accuracy, and has certain robustness on illumination variation, influence of rain and mist, indistinct license plates, broken characters and other bad conditions.

A golf swing-measuring system including a computer (16) for capturing a colored moving image obtained by photographing a golfer (11) who swings by gripping a golf club (13). The computer (16) selects and extracts one or more images each showing a swing posture as check-point images such as an address image, a take-back shaft 8 o'clock image, a take-back shaft 9 o'clock image, a take-back unskillful arm horizontal image, a top image, a downswing unskillful arm horizontal image, a downswing shaft 9 o'clock image, an impact image, a follow-through shaft 3 o'clock image, and a finish image and extracts from a large number of still images constituting said color moving image. Thereafter the computer (16) obtains coordinates of positions of attention-focused points that operate in said check-point images when said golfer swings.

The invention relates to an infrared road sign for indoor mobile robot positioning and a using method thereof. The infrared road sign consists of position identification marking points and identity identification marking points; the position identification marking points consist of the marking points at the positions of the four vertexes of a smallest outsourcing quadrilateral of all marking points, and the size of one of the four marking points which form the position identification marking points is greater than those of the other three marking points; the identify identification marking points consist of the other marking points besides the position marking points. The invention also discloses the using method for the infrared road sign for indoor mobile robot positioning. According to the infrared road sign for indoor mobile robot positioning and the using method thereof, the infrared road sign is taken as the basis of visual positioning, so that the influence of visible light illuminating in environment can be effectively reduced, the efficiency of an image processing algorithm is improved, and the visual positioning stability is improved. The four marking points are taken as the position identification marking points, so that the robustness of the algorithm is improved, and the flexibility is higher.

The present invention relates to a control method for the localization and navigation of a mobile robot and a mobile robot using the same. More specifically, the localization and navigation of a mobile robot are controlled using inertial sensors and images, wherein local direction descriptors are employed, the mobile robot is changed in the driving mode thereof according to the conditions of the mobile robot, and errors in localization may be minimized.

The embodiment of the invention provides an acoustic model optimization method and device, a voice awakening method and device, as well as a terminal. The acoustic model optimization method comprises the steps: receiving a voice signal and extracting acoustic feature parameters of the voice signal; judging whether the acoustic feature parameters correspond to a set user; when the acoustic feature parameters correspond to the set user, using the acoustic feature parameters to correct statistical parameters of a current acoustic model, so as to optimize the current acoustic model. By adopting the acoustic model optimization method provided by the embodiment of the invention, the optimized acoustic model can perform self-adaption with the specific user.

The invention applies to the technical field of intelligent equipment, and provides a proximity sensor threshold adjusting method and device and an intelligent device. The method comprises the following steps: detecting the working state of an intelligent device; starting a proximity sensor according to the working state, and adjusting the proximity threshold and leave-away threshold of the proximity sensor; performing an operation instruction corresponding to the proximity threshold or the leave-away threshold if the detected output value of the proximity sensor is more than the proximity threshold or less than the leave-away threshold, wherein the proximity value is more than the leave-away threshold, and the proximity threshold and the leave-away threshold are respectively corresponding to the working state of the intelligent device. With the adoption of the method, the proximity threshold and the leave-away threshold of the proximity sensor can be dynamically adjusted according to different working states of the intelligent device, so that the accuracy of intelligent motion recognizing in different application scenes can be increased, and the probability of inaccurate or invalid intelligent motion recognizing can be reduced.

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.

The invention discloses a cross-language end-to-end speech recognition method for the low resource Tujia language. The method comprises the following steps: preprocessing the Tujia language data; constructing a cross-language Tujia language corpus; establishing a unified coding dictionary of Chinese international phonetic alphabets and national international phonetic alphabets; establishing a cross-language end-to-end Tujia speech recognition model; and using a join temporal classification model and performing decoding under the action of the coding dictionary so as to obtain the recognition result. The recognition model with higher generalization is constructed by taking advantage of sufficient major language data and combining the idea of transfer learning so as to improve the accuracy of Tujia language speech recognition.

The invention relates to mobile phone voice control technology and discloses a voice control method of functions of a mobile phone, which is realized by voice recognition technology. The method comprises the following steps that: a user inputs selected voice key words into the mobile phone and setting the mapping relationship between the input voice key words and the functions of the mobile phone to form a voice control instruction collection; and in a voice control mode, the mobile phone automatically executes a function corresponding to a voice command when the voice command captured by the mobile phone is the same as a voice command preset by the user.

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%.

The invention discloses a multi-data fusion forest fire intelligent recognition system. The system includes a thermal infrared imager, a digital camera, an ultrasonic meteorological station, a geographic information system and an upper computer, wherein the thermal infrared imager, the digital camera, the ultrasonic meteorological station and the geographic information system are connected with the upper computer; and the upper computer is provided with an artificial neural network and a fuzzy expert system, the artificial neural system analyzes infrared images collected by the thermal infrared imager and visible light images collected by the digital camera and transmits analyzed data to the fuzzy expert system for calculation of the probability of a forest fire, the ultrasonic meteorological station provides meteorological data of a forest, and the geographic information system provides geographic information. The invention provides a multi-data fusion forest fire intelligent recognition system and method, and the method fuses various data to make forest fire predictions, and can remarkably reduce a false recognition rate of forest fires, improve operating efficiency of the system, and reduce labor intensity of workers.

The invention relates to method in technical filed of image processing, in particular to a binarization method for image processing in complex background, wherein the method comprises the following steps: carrying out edge detection of multi-scale canny operator to input image; marking connected components; drawing external rectangular frame to every independent connected component; carrying out filtering of the classifier to every rectangular frame to remove the characteristic area without character; screening out the rectangular frame with the character; extracting foreground color and background color to the rectangular frame which is screened out; carrying out solely binarization processing to every rectangular frame based on these two colors to obtain the final binarization result comprehensively. With very high accuracy and definition, the invention can greatly increase the recognition rate of the character, and the effect is superior to other methods.

A mobile phone voice key is composed of voice receiving unit, voice preprocessing module, voice key training module and unlocking module. It features that neural network voice identification technique is applied on voice key to utilize human voice to unlock key of mobile phone and to finalize status certification course of owner.

The invention discloses an 'internet+' package bottle intelligent recovery system and method. The system comprises an intelligent recovery device, user terminal equipment and a cloud service which are connected through internet communication; the intelligent recovery device comprises a human-computer interactive system, an electromechanical control device, a package bottle transmitting and rotating device, a package bottle identifying device, a package bottle compressing and crushing device, and a machine cabinet. A user selects to start throwing a bottle through the human-computer interactive system or the user terminal equipment; the intelligent recovery device has the functions of identifying the bar code, the weight, the image and the material of the package bottle. If the intelligent recovery device fails to identify, the package bottle is rejected. If the package bottle is identified successfully, the package bottle is compressed and crushed, and is recovered and stored; meanwhile, the user gets rebate. The 'internet+' package bottle intelligent recovery system is convenient to operate and use; the interaction and data sharing of the user, the intelligent recovery device, internet cloud and a user terminal are realized by utilizing the internet; the popularization of an 'internet+' resource recycling development is strongly promoted.

The invention is to provide a fingerprint scanning method and device and a gesture recognition method and device, and is mainly to solve the problems of low fingerprint scanning speed, low point reporting rate of gesture recognition and high error recognition rate in the prior art. The fingerprint scanning method comprises carrying out primary scanning on a sensor area to position a fingerprint area; and carrying out secondary scanning on the fingerprint area to obtain a fingerprint image. The fingerprint scanning method comprises two scanning; one scanning is coarse scanning carried out in the whole sensing image range, and fingerprint area images can be obtained quickly; and then, the fingerprint area is subjected to secondary scanning to obtain the fingerprint images without carrying out fine scanning on the whole sensing images, thereby ensuring precise fingerprint images and meanwhile, improving scanning speed effectively, and reducing average power consumption of sensors.

The invention discloses a fiber Bragg grating perimeter intrusion pattern recognition method and system. The method comprises the following steps: an external vibration signal which is demodulated through a fiber Bragg grating demodulation instrument is received; when the vibration amplitude of the external vibration signal is greater than a threshold value, whether an abnormal event happens is judged according to preset conditions; if the abnormal event happens, a signal with a fixed time span is intercepted from the external vibration signal, and the intercepted signal with the fixed time span is preprocessed so as to reduce the effects of system noise on a system; a time-domain signal fluctuation extreme value of the preprocessed signal is calculated; the time-domain signal fluctuation extreme value is subjected to curve fitting, and characteristic parameters of a curve after the fitting are extracted; whether the values of the characteristic parameters meet animal touch action or knock action pattern recognition conditions which are set in advance is judged; if the values of the characteristic parameters meet the conditions, pattern recognition information is generated and output to monitoring equipment. The fiber Bragg grating perimeter intrusion pattern recognition method and system can effectively recognize perimeter animal touch events, and is high in sensitivity, strong in resistance to the environment, and low in erroneous recognition rate.

The invention provides a quick acoustic event detection method under a vehicle-driving noise environment, belongs to the field of acoustic event detection of an unmanned vehicle, and is to solve the problem that an existing unmanned vehicle cannot sense sound. The method comprises the following steps: in the training stage, serving audio signals of various target acoustic events and vehicle-driving noise signals as training data, establishing an SVM model for each target acoustic event by utilizing the audio signals, and training a BPNN noise model by utilizing the obtained energy of the vehicle-driving noise signals; and in the identification stage, carrying out filtering on the audio signals of a target acoustic event to be detected collected in real time under the vehicle-driving noise environment, carrying out noise reduction and enhancement by utilizing the BPNN noise model, then, extracting MFCC coefficients, carrying out classification identification according to the MFCC coefficients and by utilizing the SVM model to determine the category of the target acoustic event to be detected, furthermore, determining corresponding whistle language sequence codes, and carrying out retrieval on a current whistle sequence library to determine corresponding whistle language information. The acoustic event detection method is used for the unmanned vehicle.

The invention is applicable to the electronic field and provides a target identification method and a device. The method comprises the following steps of: judging whether a starting action exists in ROI (region-of-interest) of a back depth map according to variation of depth value of a front depth map and the back depth map which are adjacent in a depth frame sequence; detecting regions which have the same colour frames according to a preset limb target model, and judging regions which accord with the limb target model to be limb target regions; storing characteristic set parameters of the limb target regions; tracking regions of the previous colour frame which are judged to be the limb target regions in a depth frame, and detecting the regions which are the same in the colour frame corresponding to the preset colour frame by utilizing the preset limb target model and characteristic set parameters of the previous limb target region which are stored, so as to obtain the limb target regions; acquiring coordinates of each limb target region, and identifying a target action according to the acquired coordinate values. In the target identification method and device provided by the embodiment of the invention, a depth image sequence and a colour image sequence are used for detecting the limb target regions, thus detection accuracy is effectively improved.

This invention discloses one font and sound conversion device and to one integration technique combined with sound matching, which comprises the following parts: handwriting screen, front power amplifier, A / D converter, main control system, A / D converter, back power amplifier, audio output, wherein, the main control system and memory system are connected; the sound integration module 5B and sound integration module are dispatched to realize font sound conversion.