公开(公告)号：US20220345340A1

公开(公告)日：2022-10-27

申请号：US17162056

申请日：2021-01-29

Applicant: THE REGENTS OF THE UNIVERSITY OF MICHIGAN

Inventor： Hun-Seok KIM ,  Chin-Wei HSU ,  David T. BLAAUW ,  Benjamin KEMPKE

IPC: H04L27/00

Abstract: When the ultra-low power mm-scale sensor node does not have a crystal oscillator and phase-lock loop, it inevitably exhibits significant carrier frequency offset (CFO) and sampling frequency offset (SFO) with respect to the reference frequencies in the gateway. This disclosure enables efficient real-time calculation of accurate SFO and CFO at the gateway, thus the ultra-low power mm-scale sensor node can be realized without a costly and bulky clock reference crystal and also power-hungry phase lock loop. In the proposed system, the crystal-less sensor starts transmission with repetitive RF pulses with a constant interval, followed by the data payload using pulse-position modulation (PPM). A proposed algorithm uses a two-dimensional (2D) fast Fourier transform (FFT) based process that identifies the SFO and CFO at the same time to establish successful wireless communication between the gateway and crystal-less sensor nodes.

公开(公告)号：US20180089838A1

公开(公告)日：2018-03-29

申请号：US15279581

申请日：2016-09-29

Applicant: The Regents of the University of Michigan

Inventor： Ziyun LI ,  Hun-Seok KIM

IPC: G06T7/20 ,  G06T7/60

Abstract: Optical flow is measured between a first image and a second image by evaluating a match quantifying parameter in respect of a set of candidate flow vectors. The set of candidate flow vectors includes one or more flow vectors selected in dependence upon one or more neighbor flow vectors associated with one or more neighboring pixels to the given pixel which have previously calculated respective match quantifying parameters indicative of closest matches for the one or more neighboring pixels. The set of candidate flow vectors also includes adjacent flow vectors corresponding to target pixels surrounding the target pixels identified by the neighbor flow vectors. One or more randomly selected random flow vectors is also added to the set of candidate flow vectors. The calculated match quantifying parameters are weighted in dependence upon whether the corresponding candidate flow vector is similar to any other candidate flow vector

公开(公告)号：US20230244746A1

公开(公告)日：2023-08-03

申请号：US17674470

申请日：2022-02-17

Applicant: THE REGENTS OF THE UNIVERSITY OF MICHIGAN

Inventor： Dennis SYLVESTER ,  David BLAAUW ,  Yu CHEN ,  Pierre ABILLAMA ,  Hun-Seok KIM

IPC: G06F17/14 ,  G06F17/16 ,  G06N3/08

CPC classification number: G06F17/145 ,  G06F17/16 ,  G06N3/08

Abstract: A computer-implemented method is presented for performing a computation with a neural network. The method includes: receiving a first input patch of data; applying a Walsh-Hadamard transform to the input patch to yield a transformed input patch in a transformed domain; computing an element-wise product of the transformed input patch and a kernel of the neural network; applying an inverse Walsh-Hadamard transform to the element-wise product to yield an intermediate matrix; and creating a first output patch from the intermediate matrix, where the size of the first output patch is smaller than the intermediate matrix.

公开(公告)号：US20210244280A1

公开(公告)日：2021-08-12

申请号：US17173976

申请日：2021-02-11

Applicant: THE REGENTS OF THE UNIVERSITY OF MICHIGAN

Inventor： David T. BLAAUW ,  Jamie PHILLIPS ,  Cynthia Anne CHESTEK ,  Taekwang JANG ,  Hun-Seok KIM ,  Dennis SYLVESTER ,  Jongyup LIM ,  Eunseong MOON ,  Michael BARROW ,  Samuel NASON ,  Julianna RICHIE ,  Paras PATEL

IPC: A61B5/00 ,  H04B10/80 ,  H04B10/114 ,  A61B5/37 ,  A61B5/384

Abstract: A mote includes an optical receiver that wirelessly receives a power and data signal in form of NIR light energy within a patient and converts the NIR light energy to an electrical signal having a supply voltage. A control module supplies the supply voltage to power devices of the mote. A clock generation circuit locks onto a target clock frequency based on the power and data signal and generates clock signals. A data recovery circuit sets parameters of one of the devices based on the power and data signal and a first clock signal. An amplifier amplifies a neuron signal detected via an electrode inserted in tissue of the patient. A chip identifier module, based on a second clock signal, generates a recorded data signal based on a mote chip identifier and the neuron signal. A driver transmits the recorded data signal via a LED or a RF transmitter.

公开(公告)号：US20190288887A1

公开(公告)日：2019-09-19

申请号：US15919510

申请日：2018-03-13

Applicant: The Regents of The University of Michigan

Inventor： David D. WENTZLOFF ,  Hun-Seok KIM ,  Jaeho IM

IPC: H04L27/152 ,  H04B1/403 ,  H04B1/00

Abstract: An ultra-low power back-channel receiver is presented that demodulates binary a FSK back-channel signal embedded in 5.8 GHz IEEE 802.11a Wi-Fi OFDM packets. The architecture of the back-channel receiver employs a two-step down-conversion where the first mixing stage downconverts using the third harmonic of the local oscillator for power efficiency. The LP-65 nm CMOS receiver consumes 335 μW with a sensitivity of −72 dBm at a BER of 10−3 and data-rate of 31.25 kb/s. The radio uses a balun and a 250 kHz reference crystal as external components. The receiver uses a 1V supply voltage for analog blocks, and 0.85V for digital blocks including the local oscillator and the frequency-locked loop circuits.

公开(公告)号：US20240028900A1

公开(公告)日：2024-01-25

申请号：US17872715

申请日：2022-07-25

Applicant: THE REGENTS OF THE UNIVERSITY OF MICHIGAN

Inventor： Hun-Seok KIM ,  David BLAAUW ,  Dennis SYLVESTER ,  Yu CHEN ,  Pierre ABILLAMA ,  Hyochan AN

IPC: G06N3/08 ,  G06F17/16 ,  G06F7/544

CPC classification number: G06N3/082 ,  G06F17/16 ,  G06F7/5443

Abstract: Recent advances in model pruning have enabled sparsity-aware deep neural network accelerators that improve the energy efficiency and performance of inference tasks. SONA, a novel transform-domain neural network accelerator is introduced in which convolution operations are replaced by element-wise multiplications and weights are orthogonally structured to be sparse. SONA employs an output stationary dataflow coupled with an energy-efficient memory organization to reduce the overhead of sparse-orthogonal transform-domain kernels that are concurrently processed while maintaining full multiply-and-accumulate (MAC) array utilization without any conflicts. Weights in SONA are non-uniformly quantized with bit-sparse canonical-signed-digit (BS-CSD) representations to reduce multiplications to simpler additions.

公开(公告)号：US20200116817A1

公开(公告)日：2020-04-16

申请号：US16654547

申请日：2019-10-16

Applicant: THE REGENTS OF THE UNIVERSITY OF MICHIGAN

Inventor： Li-Xuan CHUO ,  Hun-Seok KIM ,  David T. BLAAUW ,  Dennis SYLVESTER ,  Mingyu YANG

IPC: G01S5/02 ,  H04L27/26 ,  H04W64/00 ,  H04L5/14 ,  G01S11/08

Abstract: A system is presented for non-line-of-sight localization between RF enabled devices. A transmitting node is configured to transmit an RF ranging signal at a first carrier frequency, where the RF ranging signal is modulated with a symbol. The reflecting node is configured to receive the RF ranging signal and further operates to convert the RF ranging signal to a second carrier frequency and retransmit the converted ranging signal while simultaneously receiving the RF ranging signal. The localizing node is configured to receive the converted ranging signal from the reflecting node. The localizing node operates to identify, in frequency domain, the symbol in the converted ranging signal and compute a distance between the reflecting node and the localizing node based in part on the identified symbol in the converted ranging signal. The transmitting node and the localizing node may be on the same or different devices.