Abstract: A computer-implemented method (400), performed by a network node (120), for determining a candidate beam for a beam transition by a wireless communications device (100) in a beam grid, comprises: monitoring (302) beam tracking data from a plurality of wireless communications devices (110) at a plurality of locations in the beam grid over a period of time; generating (316) a beam grid history comprising monitored beam tracking data; and determining (418), based on at least the beam grid history and current beam measurement data received from a wireless communications device (110) at a location in the beam grid, at least one candidate beam for a beam transition by the wireless communications device (110).

Abstract: Methods and apparatuses herein improve bottleneck-layer processing in neural networks. Example advantages include reducing the number of accesses needed to external memory, allowing processing to run in parallel in successive bottleneck layers, based on the use of partial convolutional results, and balancing the amount of “local” memory used for storing convolutional results against the computational overhead of recomputing partial results. One aspect of the methods and apparatuses involves co-locating arithmetic and logical operators and temporary storage in the same data path, with the approach yielding both higher performance and greater energy efficiency in the implementation of bottleneck layers for neural network processing.

Abstract: An object detection arrangement (100) comprising a controller (101) configured to detect objects utilizing a multi-scale convolutional neural network, wherein the controller (101) is further configured to: receive (312, 410) image data representing an image (10) comprising an object to be detected (11) being at a distance (d) into the image (10); classify (314, 430) whether the object to be detected (11) is at risk of being incorrectly detected based on the distance (d); and if so compensate (315, 440) the object detection by shifting (317) the image (10).

Abstract: An object detection arrangement (100) comprising a controller (101) configured to detect objects utilizing a multi-scale convolutional neural network, wherein the controller (101) is further configured to: receive (312, 410) image data representing an image (10) comprising an object to be detected (11) being at a distance (d) into the image (10); classify (314, 430) whether the object to be detected (11) is at risk of being incorrectly detected based on the distance (d); and if so adapt (315, 440) the object detection accordingly.

Abstract: A method for controlling projection of an image in an extended reality device is disclosed. The extended reality device comprising an optical element, a display, and a motion tracker, wherein the optical element and/or the display are mounted movement flexible in the extended reality device. The method comprises obtaining motion information of the extended reality device from the motion tracker; determining a compensating movement of the optical element and/or the display for projection of the image based on the obtained motion information of the extended reality device; and controlling projection of the image by moving the optical element and/or the display based on the determined compensating movement. Corresponding computer program product, apparatus, and extended reality headset are also disclosed.

Abstract: A computer-implemented method (400), performed by a network node (120), for determining a candidate beam for a beam transition by a wireless communications device (100) in a beam grid, comprises: monitoring (302) beam tracking data from a plurality of wireless communications devices (110) at a plurality of locations in the beam grid over a period of time; generating (316) a beam grid history comprising monitored beam tracking data; and determining (418), based on at least the beam grid history and current beam measurement data received from a wireless communications device (110) at a location in the beam grid, at least one candidate beam for a beam transition by the wireless communications device (110).

Abstract: A method for controlling projection of an image in an extended reality device is disclosed. The extended reality device comprising an optical element, a display, and a motion tracker, wherein the optical element and/or the display are mounted movement flexible in the extended reality device. The method comprises obtaining motion information of the extended reality device from the motion tracker; determining a compensating movement of the optical element and/or the display for projection of the image based on the obtained motion information of the extended reality device; and controlling projection of the image by moving the optical element and/or the display based on the determined compensating movement. Corresponding computer program product, apparatus, and extended reality headset are also disclosed.

Abstract: An object detection arrangement (100) comprising a controller (101) configured to detect objects utilizing a multi-scale convolutional neural network, wherein the controller (101) is further configured to: receive (312, 410) image data representing an image (10) comprising an object to be detected (11) being at a distance (d) into the image (10); classify (314, 430) whether the object to be detected (11) is at risk of being incorrectly detected based on the distance (d); and if so compensate (315, 440) the object detection by adapting (316) object detection parameters.

Abstract: Methods and apparatuses herein improve bottleneck-layer processing in neural networks. Example advantages include reducing the number of accesses needed to external memory, allowing processing to run in parallel in successive bottleneck layers, based on the use of partial convolutional results, and balancing the amount of “local” memory used for storing convolutional results against the computational overhead of recomputing partial results. One aspect of the methods and apparatuses involves co-locating arithmetic and logical operators and temporary storage in the same data path, with the approach yielding both higher performance and greater energy efficiency in the implementation of bottleneck layers for neural network processing.

Abstract: An object detection arrangement (100) comprising a controller (101) configured to detect objects utilizing a multi-scale convolutional neural network, wherein the controller (101) is further configured to: receive (312, 410) image data representing an image (10) comprising an object to be detected (11) being at a distance (d) into the image (10); classify (314, 430) whether the object to be detected (11) is at risk of being incorrectly detected based on the distance (d); and if so compensate (315, 440) the object detection by shifting (317) the image (10).

Abstract: An object detection arrangement (100) comprising a controller (101) configured to detect objects utilizing a multi-scale convolutional neural network, wherein the controller (101) is further configured to: receive (312, 410) image data representing an image (10) comprising an object to be detected (11) being at a distance (d) into the image (10); classify (314, 430) whether the object to be detected (11) is at risk of being incorrectly detected based on the distance (d); and if so adapt (315, 440) the object detection accordingly.

Abstract: A method for dynamic load distribution for a distributed neural network is disclosed. The method comprises estimating, in a device of the neural network, an energy usage for processing at least one non-processed layer in the device, and estimating, in the device of the neural network, an energy usage for transmitting layer output of at least one processed layer to a cloud service of the neural network for processing. The method further comprises comparing, in the device of the neural network, the estimated energy usage for processing the at least one non-processed layer in the device with the estimated energy usage for transmitting the layer output of the at least one processed layer to the cloud service.

Abstract: Arithmetic coding utilizes probability values associated with contexts and context indexed values. The probability values are stored within a random access memory 6 from where they are fetched to a cache memory 8 before being supplied to an arithmetic encoder and decoder 4. The context indexed values used are mapped to the plurality of contexts employed such that context indexed values used to process data values close by in a position within the stream of data values being processed have a greater statistical likelihood of sharing a group of contexts than context values used to process data values far away in position within the stream of data values. Thus, a group of contexts for which the probability values are fetched together into the cache memory 8 will have an increased statistical likelihood of being used together in close proximity in processing the stream of data values. This reduces the number of cache flush operations and cache line fill operations.

Abstract: A receiver (10) for extracting a desired signal component (70) from a received signal, which received signal contains the desired signal component (70) and may contain an interfering signal component (72), wherein the receiver (10) comprises a first mixer (16) which generates an intermediate frequency signal containing a frequency-shifted version of the received signal and a frequency-shifted version of an image of the received signal, the receiver further comprising: a second mixer (26) for shifting the desired signal component (70) of the intermediate frequency signal such that it is centered on a baseband frequency, so as to generate a first composite signal containing the shifted desired signal component (70) and a shifted version of any interfering signal component (76) contained in the image of the received signal; a third mixer (30) for shifting an image of the desired signal such that it is centered on the baseband frequency, so as to generate a second composite signal containing the shifted image (74

Abstract: Arithmetic coding utilises probability values associated with contexts and context indexed values. The probability values are stored within a random access memory 6 from where they are fetched to a cache memory 8 before being supplied to an arithmetic encoder and decoder 4. The context indexed values used are mapped to the plurality of contexts employed such that context indexed values used to process data values close by in a position within the stream of data values being processed have a greater statistical likelihood of sharing a group of contexts than context values used to process data values far away in position within the stream of data values. Thus, a group of contexts for which the probability values are fetched together into the cache memory 8 will have an increased statistical likelihood of being used together in close proximity in processing the stream of data values. This reduces the number of cache flush operations and cache line fill operations.

Abstract: A receiver for isolating a wanted signal in a received signal, the receiver comprising a downconverter for downconverting the received signal in frequency to produce a downconverted signal, a filter with a passband intended for isolating that part of the spectrum of the downconverted signal that contains the wanted signal and a controller that seeks to avoid or reduce the effect of passband intrusion in the form of a negative frequency representation of an interferer, appearing in the spectrum of the received signal, upconverted in frequency to the passband. The invention consists in corresponding methods also.

Abstract: A receiver (10) for extracting a desired signal component (70) from a received signal, which received signal contains the desired signal component (70) and may contain an interfering signal component (72), wherein the receiver (10) comprises a first mixer (16) which generates an intermediate frequency signal containing a frequency-shifted version of the received signal and a frequency-shifted version of an image of the received signal, the receiver further comprising: a second mixer (26) for shifting the desired signal component (70) of the intermediate frequency signal such that it is centred on a baseband frequency, so as to generate a first composite signal containing the shifted desired signal component (70) and a shifted version of any interfering signal component (76) contained in the image of the received signal; a third mixer (30) for shifting an image of the desired signal such that it is centred on the baseband frequency, so as to generate a second composite signal containing the shifted image (74)

Abstract: Method and device for generating an address value for addressing an interleaver memory. Consecutive address fragments to which a most significant bit(s) is to be appended are generated. Only a fraction of the address fragments generated, which potentially will exceed a maximum allowable value, is compared to the maximum allowable value. If the compared address fragment exceeds the maximum allowable value it is discarded. If the compared address fragment does not exceed the maximum allowable value it is accepted.

Abstract: A receiver for isolating a wanted signal in a received signal, the receiver comprising a downconverter for downconverting the received signal in frequency to produce a downconverted signal, a filter with a passband intended for isolating that part of the spectrum of the downconverted signal that contains the wanted signal and a controller that seeks to avoid or reduce the effect of passband intrusion in the form of a negative frequency representation of an interferer, appearing in the spectrum of the received signal, upconverted in frequency to the passband. The invention consists in corresponding methods also.

Abstract: A Decimation In Frequency (DIF) Fast Fourier Transform (FFT) stage is used in an N bin FFT, wherein N is an even integer. The DIF FFT stage includes swap logic that receives a first input sample, x(v), and a second input sample, x(v+N/2), and selectively supplies either the first and second input samples at respective first and second swap logic output ports or alternatively the second and first input samples at the respective first and second swap logic output ports, wherein 0?v<N/2. The DIF FFT stage further includes a summing unit for adding values supplied by the first and second swap logic output ports; a differencing unit for subtracting values supplied by the first and second swap logic output ports; and twiddle factor logic that multiplies a value supplied by the differencing unit by a twiddle factor, WN(v+s)mod(N/2), where s is an integer representing an amount of circular shift of N input samples.