Abstract: A neural network batch normalization optimization method includes: setting a first network layer in a neural network as a starting layer; sequentially obtaining initial bias values of different network layers backwards starting from the starting layer; calculating equivalent bias values of the different network layers; determining whether there is a target network layer, wherein a ratio of the equivalent bias value corresponding to a previous layer of a target network layer to the equivalent bias value corresponding to the target network layer is no less than a pre-set threshold value; and if the target network layer is present, setting the bias values of the different network layers between the starting layer and the previous layer of the target network layer to zero, and taking the equivalent bias value of the target network layer as a bias value of the target network layer.

Abstract: A neural network batch normalization optimization method includes: setting a first network layer in a neural network as a starting layer; sequentially obtaining initial bias values of different network layers backwards starting from the starting layer; calculating equivalent bias values of the different network layers; determining whether there is a target network layer, wherein a ratio of the equivalent bias value corresponding to a previous layer of a target network layer to the equivalent bias value corresponding to the target network layer is no less than a pre-set threshold value; and if the target network layer is present, setting the bias values of the different network layers between the starting layer and the previous layer of the target network layer to zero, and taking the equivalent bias value of the target network layer as a bias value of the target network layer.

Abstract: Disclosed are a multipath selection method and device, and a storage medium. The method includes that: a correlation sequence between a received signal and a local reference signal is acquired by means of a correlation calculation method; a power spectrum of the correlation sequence and an average noise power of the received signal are acquired; according to the average noise power of the received signal, the power spectrum of the correlation sequence is divided into at least one first multipath component area and a second multipath component area according to a pre-set dividing rule; the at least one first multipath component area is searched according to a pre-set first noise threshold, so as to acquire a valid multipath component signal in the at least one first multipath component area; and the second multipath component area is searched according to a pre-set second noise threshold, so as to acquire a valid multipath component signal in the second multipath component area.

Abstract: Disclosed are a multipath selection method and device, and a storage medium. The method includes that: a correlation sequence between a received signal and a local reference signal is acquired by means of a correlation calculation method; a power spectrum of the correlation sequence and an average noise power of the received signal are acquired; according to the average noise power of the received signal, the power spectrum of the correlation sequence is divided into at least one first multipath component area and a second multipath component area according to a pre-set dividing rule; the at least one first multipath component area is searched according to a pre-set first noise threshold, so as to acquire a valid multipath component signal in the at least one first multipath component area; and the second multipath component area is searched according to a pre-set second noise threshold, so as to acquire a valid multipath component signal in the second multipath component area.

Abstract: A device for implementing FFT/DFT is disclosed, comprising: a first multiplier, a second multiplier, a first adder, a second adder, a first multiplexer, a second multiplexer, a first accumulator register, a second accumulator register and a negation controller, wherein the first adder is configured to accumulate the output signals of the first multiplexer, the first multiplier and the second multiplier and input an accumulated signal to the first accumulator register; the second adder is configured to accumulate the output signals of the second multiplexer, the first multiplier and the second multiplier and input an accumulated signal to the second accumulator register; the first accumulator register is configured to output the output signal of the first adder or feed the same back to the first multiplexer; and the second accumulator register is configured to output the output signal of the second adder or feed the same back to the second multiplexer.

Abstract: A device for implementing FFT/DFT is disclosed, comprising: a first multiplier, a second multiplier, a first adder, a second adder, a first multiplexer, a second multiplexer, a first accumulator register, a second accumulator register and a negation controller, wherein the first adder is configured to accumulate the output signals of the first multiplexer, the first multiplier and the second multiplier and input an accumulated signal to the first accumulator register; the second adder is configured to accumulate the output signals of the second multiplexer, the first multiplier and the second multiplier and input an accumulated signal to the second accumulator register; the first accumulator register is configured to output the output signal of the first adder or feed the same back to the first multiplexer; and the second accumulator register is configured to output the output signal of the second adder or feed the same back to the second multiplexer.

Abstract: The present invention discloses a channel estimation method and device in an orthogonal frequency division multiplexing system. The method includes steps of: A. grouping physical resource blocks in the bandwidth of the orthogonal frequency division multiplexing system; B. extracting at least one group from groups for channel estimation to acquire a channel coefficient, wherein the number of the extracted groups is less than the total number of groups; C. completing the MIMO demodulation by using the acquired channel coefficient; D. judging whether the channel estimation on the grouping has been completed, if yes, ending, otherwise, returning to step B. The device divides the physical resource blocks (PRB) in the bandwidth of the orthogonal frequency division multiplexing system into several groups and then carries out channel estimation processing on each group of resource blocks successively and individually, so as to achieve memory sharing and save storage amount.

Abstract: An information detection method and apparatus for a High Speed Downlink Shared Control Channel (HS-SCCH) are provided by the present disclosure, implementing the detection of HS-SCCH part1 with low false alarm probability and low false dismissal probability under the arbitrary combination of User Identifiers (UEID).

Abstract: The present invention discloses a channel estimation method and device in an orthogonal frequency division multiplexing system. The method includes steps of: A. grouping physical resource blocks in the bandwidth of the orthogonal frequency division multiplexing system; B. extracting at least one group from groups for channel estimation to acquire a channel coefficient, wherein the number of the extracted groups is less than the total number of groups; C. completing the MIMO demodulation by using the acquired channel coefficient; D. judging whether the channel estimation on the grouping has been completed, if yes, ending, otherwise, returning to step B. The device divides the physical resource blocks (PRB) in the bandwidth of the orthogonal frequency division multiplexing system into several groups and then carries out channel estimation processing on each group of resource blocks successively and individually, so as to achieve memory sharing and save storage amount.

Abstract: An information detection method and apparatus for a High Speed Downlink Shared Control Channel (HS-SCCH) are provided by the present disclosure, implementing the detection of HS-SCCH part1 with low false alarm probability and low false dismissal probability under the arbitrary combination of User Identifiers (UEID).