公开(公告)号：US20200349435A1

公开(公告)日：2020-11-05

申请号：US16934685

申请日：2020-07-21

Applicant: Massachusetts Institute of Technology

Inventor： Otkrist Gupta ,  Ramesh Raskar

IPC: G06N3/08 ,  G06N3/04 ,  G06N20/00

Abstract: A deep neural network may be trained on the data of one or more entities, also know as Alices. An outside computing entity, also known as a Bob, may assist in these computations, without receiving access to Alices' data. Data privacy may be preserved by employing a “split” neural network. The network may comprise an Alice part and a Bob part. The Alice part may comprise at least three neural layers, and the Bob part may comprise at least two neural layers. When training on data of an Alice, that Alice may input her data into the Alice part, perform forward propagation though the Alice part, and then pass output activations for the final layer of the Alice part to Bob. Bob may then forward propagate through the Bob part. Similarly, backpropagation may proceed backwards through the Bob part, and then through the Alice part of the network.

公开(公告)号：US20200146543A1

公开(公告)日：2020-05-14

申请号：US16732220

申请日：2019-12-31

Applicant: Massachusetts Institute of Technology

Inventor： Anshuman Das ,  Ramesh Raskar

IPC: A61B1/227 ,  A61B5/00 ,  G06T7/50 ,  G06T15/50 ,  G06K9/46 ,  A61B1/06

Abstract: An otoscope may project a temporal sequence of phase-shifted fringe patterns onto an eardrum, while a camera in the otoscope captures images. A computer may calculate a global component of these images. Based on this global component, the computer may output an image of the middle ear and eardrum. This image may show middle ear structures, such as the stapes and incus. Thus, the otoscope may “see through” the eardrum to visualize the middle ear. The otoscope may project another temporal sequence of phase-shifted fringe patterns onto the eardrum, while the camera captures additional images. The computer may subtract a fraction of the global component from each of these additional images. Based on the resulting direct-component images, the computer may calculate a 3D map of the eardrum.

公开(公告)号：US10330610B2

公开(公告)日：2019-06-25

申请号：US15266195

申请日：2016-09-15

Applicant: Massachusetts Institute of Technology ,  Board of Trustees of Michigan State University

Inventor： Gregory Charvat ,  Andrew Temme ,  Micha Feigin-Almon ,  Ramesh Raskar ,  Hisham Bedri

IPC: G01N22/00 ,  G01S13/88 ,  G01N21/3586

Abstract: An imaging system images near-field objects with focused microwave or terahertz radiation. Multiple antennas emit microwave or terahertz radiation, such that the radiation varies in frequency over time, illuminates a near-field object, reflects from the near-field object, and travels to a passive aperture. For example, the passive aperture may comprise a dielectric lens or a parabolic reflector. The passive aperture focuses, onto a spatial region, the microwave or terahertz radiation that reflected from the near-field object. One or more antennas take measurements, in the spatial region, of the microwave or terahertz radiation that reflected from the near-field object. A computer calculates, based on the measurements, an image of the near-field object and depth information regarding the near-field object.

公开(公告)号：US20190103876A1

公开(公告)日：2019-04-04

申请号：US16025991

申请日：2018-07-02

Applicant: Massachusetts Institute of Technology

Inventor： Ayush Bhandari ,  Felix Krahmer ,  Ramesh Raskar

IPC: H03M1/12

Abstract: A self-reset ADC may take a set of temporally equidistant, modulo samples of a bandlimited, analog signal, at a sampling rate that is greater than πe samples per second, where π is Archimedes' constant and is Euler's number. The bandlimited signal may have a bandwidth of 1 Hertz and a maximum frequency of 0.5 Hertz. These conditions of sampling rate, bandwidth and maximum frequency may ensure that an estimated signal may be recovered from the set of modulo samples. This estimated signal may be equal to the bandlimited signal plus a constant. The constant may be equal to an integer multiple of the modulus of the centered modulo operation employed to take the modulo samples.

公开(公告)号：US20190050669A1

公开(公告)日：2019-02-14

申请号：US16059144

申请日：2018-08-09

Applicant: Massachusetts Institute of Technology ,  Georgia Tech Research Corporation

Inventor： Barmak Heshmat Dehkordi ,  Albert Redo-Sanchez ,  Ramesh Raskar ,  Alireza Aghasi ,  Justin Romberg

IPC: G06K9/46 ,  G06K9/20 ,  G06K9/00

CPC classification number: G06K9/4642 ,  G06K9/00543 ,  G06K9/2018 ,  G06K9/522 ,  G06K2209/01

Abstract: A sensor may measure light reflecting from a multi-layered object at different times. A digital time-domain signal may encode the measurements. Peaks in the signal may be identified. Each identified peak may correspond to a layer in the object. For each identified peak, a short time window may be selected, such that the time window includes a time at which the identified peak occurs. A discrete Fourier transform of that window of the signal may be computed. A frequency frame may be computed for each frequency in a set of frequencies in the transform. Kurtosis for each frequency frame may be computed. A set of high kurtosis frequency frames may be averaged, on a pixel-by-pixel basis, to produce a frequency image. Text characters that are printed on a layer of the object may be recognized in the frequency image, even though the layer is occluded.

公开(公告)号：US10191154B2

公开(公告)日：2019-01-29

申请号：US15431713

申请日：2017-02-13

Applicant: Massachusetts Institute of Technology

Inventor： Achuta Kadambi ,  James Schiel ,  Ayush Bhandari ,  Ramesh Raskar ,  Vage Taamazyan

IPC: G01S17/89 ,  G01S7/48 ,  G01S17/32 ,  G01S17/10 ,  G01S7/491

Abstract: In some implementations, scene depth is extracted from dual frequency of a cross-correlation signal. A camera may illuminate a scene with amplitude-modulated light, sweeping the modulation frequency. For each modulation frequency in the sweep, each camera pixel may measure a cross-correlation of incident light and of a reference electrical signal. Each pixel may output a vector of cross-correlation measurements acquired by the pixel during a sweep. A computer may perform an FFT on this vector, identify a dual frequency at the second largest peak in the resulting power spectrum, and calculate scene depth as equal to a fraction, where the numerator is the speed of light times this dual frequency and the denominator is four times pi. In some cases, the two signals being cross-correlated have the same phase as each other during each cross-correlation measurement.

公开(公告)号：US10105049B2

公开(公告)日：2018-10-23

申请号：US15000032

申请日：2016-01-19

Applicant: Massachusetts Institute of Technology ,  Vincent Patalano, II

Inventor： Shantanu Sinha ,  Hyunsung Park ,  Albert Redo-Sanchez ,  Matthew Everett Lawson ,  Nickolaos Savidis ,  Pushyami Rachapudi ,  Ramesh Raskar ,  Vincent Patalano, II

IPC: A61B3/00 ,  A61B3/10 ,  A61B3/117 ,  G02B5/10 ,  A61B3/14

Abstract: A projector and one or more optical components project a light pattern that scans at least a portion of an anterior segment of an eye of a user, while one or more cameras capture images of the anterior segment. During each scan, different pixels in the projector emit light at different times, causing the light pattern to repeatedly change orientation relative to the eye and thus to illuminate multiple different cross-sections of the anterior segment. The cameras capture images of each cross-section from a total of at least two different vantage points relative to the head of the user. The position of the projector, optical components and cameras relative to the head of the user remains substantially constant throughout each entire scan.

公开(公告)号：US09894254B2

公开(公告)日：2018-02-13

申请号：US15151417

申请日：2016-05-10

Applicant: Massachusetts Institute of Technology

Inventor： Barmak Heshmat Dehkordi ,  Ik Hyun Lee ,  Hisham Bedri ,  Ramesh Raskar

IPC: H04N5/225 ,  G02B6/06

CPC classification number: H04N5/2256 ,  G02B6/06 ,  H04N5/2257

Abstract: An open-ended, incoherent bundle of optical fibers transmits light from a nearby scene. A camera captures images of the back end of the fiber bundle. Because the fiber bundle is incoherent, the captured image is shuffled, in the sense that the relative position of pixels in the image differs from the relative position of the scene regions that correspond to the pixels. Calibration is performed in order to map from the front end positions to the back-end positions of the fibers. In the calibration, pulses of light are delivered, in such a way that the time at which light reflecting from a given pulse enters a given fiber directly correlates to the position of the front end of the given fiber. A time-of-flight sensor takes measurements indicative of these time signatures. Based on the map obtained from calibration, a computer de-shuffles the image.

公开(公告)号：US20170248532A1

公开(公告)日：2017-08-31

申请号：US15517122

申请日：2015-10-29

Applicant: Massachusetts Institute of Technology ,  The General Hospital Corporation

Inventor： Achuta Kadambi ,  Ramesh Raskar ,  Rajiv Gupta ,  Adam Pan

IPC: G01N23/083 ,  H01J35/14 ,  G01T1/24 ,  H01J35/06 ,  G01N23/04 ,  G21K1/04

CPC classification number: G01N23/083 ,  G01N23/046 ,  G01N2223/04 ,  G01N2223/1016 ,  G01N2223/304 ,  G01N2223/419 ,  G01T1/248 ,  G21K1/043 ,  H01J35/065 ,  H01J35/14

Abstract: For each X-ray path through a tissue, numerous trials are conducted. In each trial, X-ray photons are emitted along the path until a Geiger-mode avalanche photodiode “clicks”. A temporal average—i.e., the average amount of time elapsed before a “click” occurs—is calculated. This temporal average is, in turn, used to estimate a causal intensity of X-ray light that passes through the tissue along the path and reaches the diode. Based on the causal intensities for multiple paths, a computer generates computed tomography (CT) images or 2D digital radiographic images. The causal intensities used to create the images are estimated from temporal statistics, and not from conventional measurements of intensity at a pixel. X-ray dosage needed for imaging is dramatically reduced as follows: a “click” of the photodiode triggers negative feedback that causes the system to halt irradiation of the tissue along a path, until the next trial begins.

公开(公告)号：US09448060B2

公开(公告)日：2016-09-20

申请号：US15004933

申请日：2016-01-23

Applicant: Massachusetts Institute of Technology

Inventor： Andreas Velten ,  Ramesh Raskar

IPC: G01B11/00 ,  G01N21/27 ,  G01B11/25 ,  G01N21/25 ,  G01N21/64 ,  G01J3/02 ,  G01J3/10 ,  G01J3/42 ,  G01N21/55 ,  G01N21/31

CPC classification number: G01B11/002 ,  G01B11/2513 ,  G01J3/021 ,  G01J3/027 ,  G01J3/10 ,  G01J3/42 ,  G01N21/255 ,  G01N21/27 ,  G01N21/55 ,  G01N21/64 ,  G01N21/6456 ,  G01N2021/3129 ,  G01N2021/6417 ,  G01N2201/061

Abstract: An active imaging system, which includes a light source and light sensor, generates structured illumination. The light sensor captures transient light response data regarding reflections of light emitted by the light source. The transient light response data is wavelength-resolved. One or more processors process the transient light response data and data regarding the structured illumination to calculate a reflectance spectra map of an occluded surface. The processors also compute a 3D geometry of the occluded surface.

Abstract translation: 包括光源和光传感器的主动成像系统产生结构化照明。 光传感器捕获关于光源发射的光的反射的瞬态光响应数据。 瞬态光响应数据经波长分辨。 一个或多个处理器处理关于结构化照明的瞬态光响应数据和数据，以计算遮挡表面的反射光谱图。 处理器还计算封闭表面的3D几何形状。