Huawei Technologies Co., Limited

Huawei Device Co., Ltd.

Hongrui Jiang

Xiyong Wang

Weijun Zheng

Junbin Liang

Qiang Ding

Junyang Zhou

Liyao Zhang

Huajun Cao

Wenmei Gao

Xiaojuan Li

Yahui Wang

Hao Jing

Li Zhou

Qinghua Yu

Lei Zhao

Zheng Lin

Zhicheng Sui

Mohan Liu

Liang Yuan

Yunbin Wang

Kunpeng Li

A neural network model processing method includes obtaining a first low-bit neural network model through training, where the model includes a first operation layer and a second operation layer. Each operation layer includes at least one operation. Values/a value of a parameter and/or data used for the operation are/is represented by using N bits, and N is a positive integer less than 8. The neural network model processing method further includes compressing the model to obtain a second low-bit neural network model, where the compressed model includes a third operation layer. The third operation layer is equivalent to the first operation layer and the second operation layer, and an operation layer other than the third operation layer in the at least one operation layer is the same as an operation layer other than the first operation layer and the second operation layer in the at least two operation layers.

Neural network model processing method and apparatus

A target monitoring method, a camera, a controller, and a target monitoring system, where the system includes a first camera and a second camera. An overlapping area exists between fields of view of the first camera and the second camera. The method includes obtaining, location information of a to-be-tracked target in a first monitoring picture when the first camera is used as a current primary monitoring camera, determining, based on the location information of the target in the first monitoring picture, whether a location of the target in the first monitoring picture is in the overlapping area, where the overlapping area is an overlapping range between the fields of view of the first camera and the second camera, and switching, the current primary monitoring camera to the second camera when the location of the target in the first monitoring picture is in the overlapping area.

Target monitoring method, camera, controller, and target monitoring system

A terminal display scanning method for optical fingerprint recognition is provided, the method includes: if a start operation on a terminal screen is detected, obtaining a first vector set, where the first vector set includes A data vectors that are mutually orthogonal or mutually quasi-orthogonal, each of the data vectors includes a plurality of data elements, and A is an integer greater than 1; sequentially using the data elements in each of the data vectors to control light emitting of minimum pixel units in a fingerprint recognition area until controlling light emitting of all the minimum pixel units in the fingerprint recognition area is completed, to obtain a second vector set corresponding to the fingerprint recognition area, where the second vector set carries fingerprint information; and demodulating the second vector set by using the first vector set to obtain all fingerprint information in the fingerprint recognition area.

Fingerprint information obtaining method and fingerprint recognition apparatus

When a location of a wearable system changes, the wearable system can ensure that a gesture working region always remains in a field of view of the wearable system. Therefore, the wearable system can still collect and recognize a gesture image and implement gesture control. The solution provided in this application is as follows: The wearable system creates a gesture working region and a region of interest, where the region of interest is within a shooting area of a first camera, and the first camera is included in a wearable device; obtains a location parameter of the wearable system; adjusts the region of interest according to the location parameter, so that the region of interest covers the gesture working region; collects a gesture image within the gesture working region; recognizes the gesture image to obtain a gesture instruction; and performs a corresponding operation according to the gesture instruction.

Gesture control method for wearable system and wearable system

A method and an apparatus for converting a speech-to-text in a multiparty call. Receiving speech-to-text requests sent by at least two terminals, where the speech-to-text requests include a first identifier and a second identifier; allocating a session to the at least two terminals, so that in the speech-to-text requests sent by the at least two terminals, terminals that have a same first identifier or a same second identifier have a same session; receiving, by using a packet-switched domain, a voice stream whose sampling rate is greater than 8 KHz and that is sent by at least one terminal in one multiparty call, where the multiparty call corresponds to one session; converting the voice stream into a text; and sending the text to a terminal in the multiparty call.

Method and apparatus for converting voice into text in multiparty call

Embodiments of the present invention relate to a method and an apparatus for optimizing a model applicable to pattern recognition, and a terminal device. The terminal device receives a universal model delivered by a server, where the universal model includes an original feature parameter; recognizes target information by using the universal model, and collects a plurality of local samples; when a model optimization condition is met, corrects the original feature parameter by using a first training algorithm to obtain a new feature parameter; and optimizes the universal model according to a second training algorithm and the new feature parameter, to obtain an optimized universal model. That is, in the present invention, the terminal device further optimizes, according to the collected local samples, the universal model received from the server to obtain a relatively personalized model applicable to pattern recognition.

Method and apparatus for optimizing model applicable to pattern recognition, and terminal device

This application relates to the field of intelligent hardware, and in particular, to an intelligent scenario in which a mobile terminal and a car are combined. After detecting, by using an acceleration sensor and a gyroscope of the mobile terminal, that a car engine stops, the mobile terminal receives, from a sensor installed near a roof lamp of the car, a signal indicating that the roof lamp is on, and determines that a door of the car is opened. In this case, the mobile terminal sends a reminder to a user.

Intelligent car reminding method and apparatus

A method is provided that includes: displaying, by the terminal, a randomly generated dynamic password on a display interface, and starting a voice receiving device of the terminal; obtaining by using the voice receiving device, voice information when a user reads the dynamic password, converting the voice information into a text password, and extracting a voiceprint feature of the voice information; and when the text password is the same as the dynamic password, determining terminal usage permission of the user according to a prestored correspondence between terminal usage permission and a voiceprint feature, and setting the display interface to an unlocked or locked state according to the determined terminal usage permission; or setting the display interface to a locked state when the text password is different from the dynamic password.

Apparatus and methods for improving terminal security

An information processing method and apparatus in the field of mobile communications technologies is provided. In the method, a first terminal obtains, during a call with a second terminal by using voice information sent by the second terminal, to-be-obtained information and object information of an object to which the to-be-obtained information belongs. The first terminal determines at least one category to which the to-be-obtained information belongs and determines a group that is in any category of the at least one category and corresponding to the object to which the to-be-obtained information belongs. The first terminal searches information in the corresponding group in any category for target information associated with the object information and processes the target information with no manual searching being performed.

Information processing method and apparatus

An image collecting method and apparatus are provided, where the image collecting method includes enabling an image collecting unit in an electronic device, when it is identified that after a first specific motion, a gesture of the electronic device changes to a specific device gesture; monitoring whether the electronic device performs a second specific motion in the specific device gesture; and when it is detected that the electronic device performs the second specific motion in the specific device gesture, controlling the image collecting unit to obtain an at least first image by means of collection within motion time in which the electronic device performs the second specific motion.

Image collecting method and apparatus

Embodiments of the present invention provide a voice identification method, which includes: obtaining voice data; obtaining a confidence value according to the voice data; obtaining a noise scenario according to the voice data; obtaining a confidence threshold corresponding to the noise scenario; and if the confidence value is greater than or equal to the confidence threshold, processing the voice data. An apparatus is also provided. The method and apparatus that flexibly adjust the confidence threshold according to the noise scenario greatly improve a voice identification rate under a noise environment.

Voice identification method and apparatus

Embodiments of the present invention provide a voice identification method, including: obtaining voice data; obtaining a first confidence value according to the voice data; obtaining a noise scenario according to the voice data; obtaining a second confidence value corresponding to the noise scenario according to the first confidence value; and if the second confidence value is greater than or equal to a pre-stored confidence threshold, processing the voice data. An apparatus is also provided. The method and apparatus that flexibly adjust the confidence value according to the noise scenario greatly improve a voice identification rate under a noise environment.

An information recognition method and apparatus are provided. The method includes receiving, by a terminal, voice information, extracting a voice feature from the voice information, performing matching calculation on the voice feature and a phoneme string corresponding to each candidate text in multiple candidate texts to obtain a recognition result, where the recognition result includes at least one command word and a label corresponding to the at least one command word, and recognizing, according to the label corresponding to the at least one command word, an operation instruction corresponding to the voice information. A terminal recognizes text information, which is corresponding to voice information input by a user, as an operation instruction.

Growing community of inventors

Shenzhen, China

Hongrui Jiang