National Taiwan University

Industrial Technology Research Institute

Mediatek Corporation

National Science Council

Via Technologies, Inc.

National Science Council of Republic of China

Acer Inc.

Tzi-Dar Chiueh

Jyh-Horng Chen

Edzer Lienson Wu

Hwai-Tsu Chang

Kuan-Hung Chen

Li-Wei Kuo

Yun-An Huang

Ke-Horng Chen

Chun-Nan Chen

Sy-Yen Kuo

Jin-Bin Yang

Yuan-Hao Huang

Sung-Chiao Li

Chung-Chih Chang

Jen-Shi Wu

Tsung-Te Liu

Chun-Yuan Chu

Chan-Shih Lin

Po-Chun Hsieh

Huang-Lin Yang

Wan-Yu Meng

Fang-Ru Hsu

Po-Chen Lin

Yi-Yao Lan

Yeh-Rong Hsu

Ying-Tuan Hsu

Bei-Sheng Su

Ching-Ting Huang

A computing-in-memory circuitry includes multiple digital-to-analog converters, multiple computing arrays, and multiple charge processing networks. The digital-to-analog converters convert external data into input data and the digital-to-analog converters are connected in series with a corresponding plurality of output capacitor pairs. The computing arrays receive the input data from both ends and execute a computation to output a first computing value. The charge processing networks receive and accumulate the first computing values over a predetermined time interval through switching pairs in series with the output capacitor pairs. The charge processing networks evenly distribute charges of the first computing value to selected output capacitor pairs and compare voltage differences between two ends of the output capacitor pairs to output a second computing value.

Computing-in-memory circuitry

A method of decoding polar codes based on belief propagation includes conventional belief propagation to decode the polar codes first; when a number of iterations exceeds a predefined upper limit and a cyclic redundancy check fails, the method selects log-likelihood ratio vectors of a plurality of R or L messages from a plurality of log-likelihood ratio vectors generated in each of the iterations and generates another set of log-likelihood ratio vectors (referred to as candidate vector group) to be used as initial values of the R or L messages for a subsequent belief propagation to perform belief propagation decoding iterations and cyclic redundancy check again. Whenever a decoding result passes the cyclic redundancy check, the method exits; otherwise, the method iterates the above procedure until a maximum number of candidate vector groups has been reached.

Method of decoding polar codes based on belief propagation

A method of training a neural network including multiple neural network weights and multiple neurons, and the method includes using floating-point signed digit numbers to represent each of the multiple neural network weights, wherein a mantissa of each of the multiple neural network weights is represented by multiple mantissa signed digit groups and an exponent of each of the multiple neural network weights is represented by an exponent digit group; and using the exponent digit group and at least one of the multiple mantissa signed digit groups to perform weight adjustment computation and neural network inference computation.

Method of neural network training using floating-point signed digit representation

A method and apparatus for single carrier wideband magnetic resonance imaging (MRI) data acquisition are provided. The method includes the following steps: exciting a slice or slab with the use of RF pulse and a slice/slab selection gradient; applying a phase encoding gradient along a phase encoding direction and reducing a FOV along the phase encoding direction by a factor of W through k-space subsampling; applying a frequency encoding gradient along a frequency encoding direction and increasing a FOV along the frequency encoding direction by a factor of W; and applying a separation gradient along the phase encoding direction during the frequency encoding duration and k-space data acquisition.

Method and apparatus for single carrier wideband magnetic resonance imaging (MRI) data acquisition

The present invention discloses a 3-dimension magnetic resonance imaging method which comprises: applying a slab selection gradient to a subject; transmitting a radiofrequency pulse to the subject, and exciting a slab of the subject to produce magnetic resonance signals with a continuous frequency bandwidth; performing a spatial encoding gradient across three dimensions to encode the magnetic resonance signals, wherein an equivalent encoded field of view which along the selected acceleration direction is controlled by the spatial encoding gradient, and the equivalent encoded field of view is shorter than the excited slab size of the subject; applying a separation gradient along with the spatial encoding gradient; and receiving and reconstructing the encoded magnetic resonance signals to produce 3D images.

Method and apparatus for 3D magnetic resonance imaging

The present invention provides a method and an apparatus for cell search and synchronization for subscriber stations of the Long Term Evolution (LTE) system. The invention uses primary synchronizing signal of primary synchronization code in each frame structure to establish synchronization with the base station when a subscriber station accesses the LTE network. With such synchronization between the subscriber station and the base station, control signals and transmission data may be correctly exchanged between them.

Method and apparatus for cell search and synchronization in mobile communication

A method and an apparatus for enhancing signals in magnetic resonance imaging are provided. The method includes the following steps. Applying one or more than one RF pulse, which carries at least two frequency components, and a slice/slab selection gradient to a subject, so that at least two slices/slabs of the subject respectively corresponding to the at least two frequency components are excited simultaneously. Applying a plurality of spatial encoding gradients and one or more than one separation gradients for separating the at least two slices/slabs. Receiving a plurality of responsive RF signals excited from the subject. The responsive RF signals are restored according to a signal restoration function.

Method and apparatus for enhancing signal in magnetic resonance imaging

A method for eliminating interference in a receiver (RX) includes: obtaining Bluetooth (BT) information and wireless local area network (Wireless LAN, WLAN) information by performing detection on at least one RX signal, wherein the RX signal is received through an antenna of the RX, and the RX signal includes components of both a BT signal and a WLAN signal that are respectively emitted from at least one transmitter (TX); and performing maximum likelihood (ML) joint detection on frequency representatives of the RX signal according to the BT information and the WLAN information and according to a predetermined library, in order to extract at least one of a reproduced version of the BT signal and a reproduced version of the WLAN signal. An associated apparatus is also provided.

Method for eliminating interference in a receiver, and associated apparatus

A method for acquiring MRI signals includes: applying one or more than one RF pulse, which carries at least two frequency components, and a slice/slab selection gradient to a subject, so that at least two slices/slabs of the subject respectively corresponding to the at least two frequency components are excited simultaneously; applying a plurality of spatial encoding gradients; applying a plurality of separation gradients for separating the at least two slices/slabs; and applying at least one coherent refocusing gradient between the plurality of separation gradients.

Method and apparatus for acquiring magnetic resonance imaging signals

A diffusion imaging method is provided. The diffusion imaging method includes performing a plurality of data collection sequences. Each data collection sequence includes applying an excitation radio frequency signal and a selection gradient. The excitation radio frequency signal includes a first set of frequency bands selected to simultaneously excite a first nuclei type in a plurality of cross sections of a subject. Each data collection sequence further includes applying a diffusion gradient during formation of a magnetic resonance signal, applying a spatial encoding gradient during formation of the magnetic resonance signal, and while acquiring the magnetic resonance signal, applying a separation gradient to change a frequency separation between portions of the magnetic resonance signal. The diffusion imaging method further includes computationally determining a diffusion image of each of the plurality of cross sections.

Simultaneous diffusion imaging of multiple cross sections

A wireless remote control system is provided. This system includes a remote controller, plural detectors, and a decision module. The detectors are used for detecting the frequency a wireless signal emitted by the remote controller and respectively generate a detecting result. Based on at least one frequency difference between the detecting results, the decision module determines how the remote controller is moving and thereby generates a control signal. Because the decision module needs no knowledge of the frequency of the emitted wireless signal, the wireless remote controller has the advantages of small size, low cost, and low power consumption.

Wireless remote control system

A magnetic resonance imaging (MRI) method is provided, including the steps of generating a wideband RF signal that has a plurality of frequency bands respectively corresponding to a plurality of different kinds of nuclei, to simultaneously excite the different kinds of nuclei, detecting a wideband responsive RF signal that has a plurality of frequency bands respectively emitted by the different kinds of nuclei, and reconstructing magnetic resonance images for the different kinds of nuclei based on the wideband responsive RF signal. An MRI apparatus employing the MRI method is also provided.

Wideband magnetic resonance imaging apparatus and method

Provided is a method for simultaneously acquiring magnetic resonance slices/slabs of a subject. The method comprises steps as follows. First, apply one or more than one RF pulse, which carries at least two frequency components, and a slice/slab selection magnetic field gradient so that at least two slices/slabs of the subject respectively corresponding to the at least two frequency components are excited simultaneously. Second, apply a spatial encoding magnetic field gradient. Third, apply a slice/slab separation magnetic field gradient so as to separate the at least two slices/slabs. The method according to the present invention can be used to acquire data for simultaneously reconstructing multiple slices/slabs. The method is compatible with existing MRI systems.

Method and apparatus for simultaneously acquiring multiple slices/slabs in magnetic resonance system

A series of digit processing units (DPUs) are connected to form a finite impulse response (FIR) filter. Each DPU includes a register, a multiplexer, and a coefficient multiplier. The register stores and delays an input digital signal to be filtered. The multiplexer has inputs connected to the input node and to an output of the register, an output of the multiplexer for connecting to a next stage DPU. The coefficient multiplier is connected to the output of the register and multiplies the input signal by a coefficient or part of a coefficient. A group of DPUs can have multiplexers set so that the register of each DPU stores the same part of the input signal for processing a single filter coefficient. An adder is provided to sum output of the DPUs and output a filtered signal. The critical path of the FIR filter is independent of coefficient number and precision.

Reconfigurable fir filter

A Viterbi/Turbo unified decoder supports both voice and data streams due to the ability of performing Viterbi (convolutional) decoding and Turbo decoding. The Viterbi/Turbo unified decoder of an embodiment reduces the hardware cost by computing path metrics for both Viterbi and Turbo decoding using a single control circuit. The control circuit comprises a plurality of processors and memory banks, and the routing rule for the processors to read/write the path metric information from/to the memory banks are fixed for both Viterbi and Turbo coded inputs.

Growing community of inventors

Taipei, Taiwan

Tzi-Dar Chiueh