Abstract: Disclosed are a method, an apparatus and an electronic device for controlling audio playback of multiple loudspeakers, wherein the method comprises: determining the location information of each speaker and voice signals issued by each speaker; determining the area where each speaker is located according to the location information of each speaker; determining voice instruction corresponding to each voice signal; and controlling the multiple loudspeakers to play the audio indicated by the corresponding voice instruction respectively for the area where the speaker of each voice instruction is located. According to the method and apparatus and/or the electronic device in an embodiment of the present disclosure, different audios can be played for different areas in a preset space.

Abstract: The present disclosure provides a method and apparatus for deconvolving feature data using convolution hardware. The method includes: reading a feature map and deconvolution kernel into on-chip memory, and padding zeroes to the feature map; determining convolution kernels based on the deconvolution kernel; removing a row and/or column of each convolution kernel whose elements all are invalid weights, to obtain an optimized convolution kernel, and removing a corresponding row and/or column in the zero-padded feature map to obtain an corresponding optimized feature map; convolving each optimized convolution kernel with corresponding optimized feature map using the multiply-add array, to obtain convolutional outputs; and interleaving and synthesizing the convolutional outputs to obtain an interleaving synthetic output including at least a deconvolutional output corresponding to the feature map and deconvolution kernel.

Abstract: A method, a device, a computer readable medium and an electronic apparatus for speech signal processing are disclosed. The method comprises: acquiring sound source position information and at least two channels of sound signals from a microphone array; suppressing, according to the sound source position information, a sound signal from the sound source direction in the at least two channels of sound signals, to obtain a noise reference signal of the microphone array; acquiring, according to the sound source position information, a sound signal from the sound source direction in the at least two channels of sound signals, to obtain a speech reference signal; removing, based on the noise reference signal, a residual noise signal in the speech reference signal to obtain a desired speech signal.

Abstract: Disclosed are a feature extraction method and apparatus for a three-dimensional feature map, a storage medium, and an electronic device. The method includes: determining an overlay parameter based on depth information of a three-dimensional feature map to be processed; decomposing the three-dimensional feature map into a plurality of target two-dimensional feature maps based on the depth information and the overlay parameter; performing two-dimensional convolution processing on each of the plurality of target two-dimensional feature maps to obtain a plurality of initial feature maps; and determining a target feature map corresponding to the three-dimensional feature map based on the plurality of initial feature maps.

Abstract: This application discloses a data processing method and apparatus, an electronic device, and a storage medium. When execution is performed at an operation layer of a neural network model, based on a pre-stored buffer allocation relationship, a first address range for cyclic addressing is set for a first buffer corresponding to input data and a second address range for cyclic addressing is set for a second buffer corresponding to an output result. Subsequently, cyclic addressing can be performed in the first buffer based on the first address range for cyclic addressing, to read the input data for the operation layer; and cyclic addressing can be performed in the second buffer based on the second address range for cyclic addressing, to write the output result of the operation layer into the second buffer. In this way, efficiency of buffer utilization can be effectively improved, and further operation efficiency for the model is improved.

Abstract: Disclosed are a dynamic display method based on an operating body, a computer-readable storage medium and electronic device. The method includes: determining a position of a to-be-displayed icon on a screen in response to an air operation performed by the operating body relative to the screen in space; displaying the icon at the determined position; determining a first movement physical quantity of the operating body in response to detecting the operating body moving; determining a second movement physical quantity of the icon on the screen based on the first movement physical quantity; and controlling the icon to be dragged on the screen based on the second movement physical quantity. Thus, a user can know whether the air operation is accurate or not in real time according to a moving track of the icon, which improves the user's sense of control and the accuracy of the air operation.

Abstract: Disclosed are a neural network training method, a neural network training device and an electronic device. The neural network training method includes: training a first neural network to be trained by using sample data; determining an indicator parameter of the first neural network in a current training process; determining an update manner corresponding to a preset condition if the indicator parameter meets the preset condition; and updating a parameter of a batch normalization layer in the first neural network based on the update manner. In this way, sparsing of a feature map output by a neural network is implemented, thereby reducing an amount of data to be transmitted and improving computation speed of a chip.

Abstract: Disclosed are a method, an apparatus and an electronic device for controlling audio playback of multiple loudspeakers, wherein the method comprises: determining the location information of each speaker and voice signals issued by each speaker; determining the area where each speaker is located according to the location information of each speaker; determining voice instruction corresponding to each voice signal; and controlling the multiple loudspeakers to play the audio indicated by the corresponding voice instruction respectively for the area where the speaker of each voice instruction is located. According to the method and apparatus and/or the electronic device in an embodiment of the present disclosure, different audios can be played for different areas in a preset space.

Abstract: A map construction method, an electronic device and a readable storage medium are disclosed. The map construction method includes: obtaining three-dimensional point cloud data corresponding to observed objects in current target frame data, the observed objects including a feature object and a non-feature object; obtaining a feature parameter corresponding to the feature object based on three-dimensional point cloud data corresponding to the feature object, the feature parameter being used to indicate a position and a size of the feature object; and adjusting a point cloud density of three-dimensional point cloud data corresponding to the non-feature object to a preset value, to construct a map according to the feature parameter and the three-dimensional point cloud data of which the point cloud density is adjusted.

Abstract: Disclosed are a method for updating a neural network and an electronic device. The method includes: inputting a first image set having tag information into a first depth neural network, and determining a cross entropy loss value of the first image set by using the first depth neural network; inputting a second image set having no tag information separately into the first depth neural network and a second depth neural network, and determining a consistency loss value of the second image set, the first depth neural network and the second depth neural network having the same network structure; updating parameters of the first depth neural network based on the cross entropy loss value and the consistency loss value; and updating parameters of the second depth neural network based on the updated parameters of the first depth neural network.

Abstract: Disclosed are a speech recognition method and device, and an electronic equipment. In the speech recognition method, when a user performs speech input, the lip of the user may be shot while audio is collected, then a second lip region of the user in a current frame image is obtained based on the current frame image and at least one first lip region in a historical frame image; concurrently, a second speech feature of current frame audio may be obtained based on current frame audio and at least one first speech feature of historical frame audio. Then, the phoneme probability distribution of the current frame audio may be obtained according to the speech features and the lip regions, and then the speech recognition result of the current frame audio may be obtained according to the phoneme probability distribution.

Abstract: A positioning method of a mobile device includes: determining a first position and orientation parameter of a mobile device when a current frame image is captured, and determining a straight line corresponding to a preset sign in the current frame image; determining a plurality of second position and orientation parameters based on the first position and orientation parameter; determining, in a high-definition map, point cloud data within a preset range of a geographic location when the current frame image is captured; converting the point cloud data within the preset range into a pixel plane-coordinate system to obtain a plurality of second image coordinate sets; determining, based on distances from image coordinates in the plurality of second image coordinate sets to the straight line, a position and orientation parameter of the mobile device when the current frame image is captured among the plurality of second position and orientation parameters.

Abstract: Disclosed are a convolution processing engine and a control method thereof, and a convolutional neural network accelerator comprising the convolution processing engine. The convolution processing engine comprises at least two cache memories connected in series and an operational circuit. The convolution processing engine can realize an efficient convolution operation with lower complexity and power consumption.

Abstract: Disclosed are a method for generating a neural network, an apparatus thereof, and an electronic device. The method includes: obtaining an optimal neural network and a worst neural network from a neural network framework by using an evolutionary algorithm; obtaining an optimized neural network from the optimal neural network by using a reinforcement learning algorithm; updating the neural network framework by adding the optimized neural network into the neural network framework and deleting the worst neural network from the neural network framework; and determining an ultimately generated neural network from the updated neural network framework. In this way, a neural network is optimized and updated from a neural network framework by combining the evolutionary algorithm and the reinforcement learning algorithm, thereby automatically generating a neural network structure rapidly and stably.

Abstract: Disclosed are a method and an apparatus for accessing tensor data. The method may include determining a first row address in a first memory where one or more first data items to be accessed in a logical structure of the tensor data are stored at the first row address, copying data items at the first row address in the first memory to a first buffer row of a first buffer, moving each first data item in the first buffer row of the first buffer to a corresponding location at least in a first buffer row of a second buffer, and storing data items in the first buffer row of the second buffer into corresponding target locations in the second memory.

Abstract: The present disclosure provides an instruction execution method, device, and electronic equipment. In the instruction execution method described above, after obtaining an exceptional signal generated by a neural network processor during an operation, the electronic equipment determines an exception processing instruction corresponding to the exceptional signal according to the exceptional signal, then it determines a first instruction queue needed to be executed by the neural network processor, and then it generates a second instruction queue based on the exception processing instruction and the first instruction queue, and finally it controls the neural network processor to execute the second instruction queue, so that errors encountered by the neural network processor can be timely processed, thereby shortening the error processing delay and improving the data processing efficiency of the hardware system in the electronic equipment.

Abstract: A method, an apparatus, and an electronic device for controlling memory access are disclosed. According to an embodiment, there is provided a method for controlling access to a memory including a plurality of memory modules configured in parallel. The method comprises: receiving an access instruction including an addressing field which comprise a parallel control field for controlling parallel access, a module address field for indicating a memory module, and an in-module address field for indicating an addresses within a memory module; parsing the access instructions to determine the parallel control field, the module address field and the in-module address field; determining one or more memory modules to be accessed based on the parallel control field and the module address field; and accessing one or more addresses which are within the one or more memory modules to be accessed and assigned by the in-module address field.

Abstract: A method, a device, a computer readable medium and an electronic apparatus for speech signal processing are disclosed. The method comprises: acquiring sound source position information and at least two channels of sound signals from a microphone array; suppressing, according to the sound source position information, a sound signal from the sound source direction in the at least two channels of sound signals, to obtain a noise reference signal of the microphone array; acquiring, according to the sound source position information, a sound signal from the sound source direction in the at least two channels of sound signals, to obtain a speech reference signal; removing, based on the noise reference signal, a residual noise signal in the speech reference signal to obtain a desired speech signal.

Abstract: An electronic device includes: a circuit board including a substrate and at least one first electronic element disposed on one side of the substrate; a first heat dissipation component; and a first heat conduction component including a first heat conduction plate and at least one first heat conduction member disposed on the first heat conduction plate and corresponding to the at least one first electronic element. A material of the first heat conduction component is different from that of the first heat dissipation component, and the first heat conduction component is disposed between the first heat dissipation component and the circuit board.