Comments on patents below: Moura's work is theoretical in nature but has also impacted industry. The work in SPIRAL has been licensed by companies like Intel (for example, Intel’s IPP library includes a module, IPPg, the g standing for generated, that includes several thousands of algorithms and many megabytes of code; Intel's MKL library also includes multiple kernels automatically generated by SPIRAL.) SPIRAL has been licensed by SPIRALGen, cofounded by Moura and CMU colleagues to develop and commercialize SPIRAL. Several of his patents have been listed as prior art by numerous patents of US and foreign companies. For example:
- Patents 6,201,839 and 6,438,180: The two patents by Kavcic and Moura (6,201,839 and 6,438,180) are listed as prior art in over one hundred patents. Many of these are by magnetic recording companies like Marvell (26 patents), IBM/Hitachi (7 patents), LSI/Agere (6 patents), Seagate/Maxtor (9 patents), Infineon (4 patents). These patents on signal dependent correlation sensitive sequence detectors provided a significant edge to the companies that adopted them in successive generations of detector chips in the last 10 years.
Signal dependent noise affects other domain applications, in particular, in cellular mobile communications. Here, the signal dependence arises because in urban environments or inside buildings, signals between source and destination bounce off walls or others structures and these multiple replicas (multipath) may interfere destructively. Again, the signal dependent correlation sensitive detector developed by Moura and Kavcic can be adapted to improve the Viterbi detector. The same two patents 6,201,839 and 6,438,180 have been cited as prior art by patents of well known communication chip design companies like Qualcomm/ Flarion (16) or Broadcom (2) among several other companies.
- Patent 5,900,778: This patent by Stonick, Stonick, and Moura proposes a novel solution to the significant distortion arising in power amplifiers. The traditional solution compensated the power amplifier highly nonlinear distortion by solving an inverse problem, which was slow to converge, introduced bias, other errors, and a long delay. The solution dsiclosed in this patent turned the problem on its head to find an elegant practical solution that predistorts the signal prior to the power amplifier by first determining the much easier forward model and then inverting the forward model; this provides an unbiased solution to the high power amplifier distortion that also converges much faster. This technique has motivated at least 27 companies to develop similar predistorters. Patent 5,900,778 has been referenced as prior art in at least 70 patents covering applications ranging from TV broadcasting (e.g., Harris Corporation, 5 patents), to wireline and wireless infrastructure providers (e.g., PMC-Sierra, 10 patents; Alacatel/Lucent, 6, Ericson, 9; Nokia, 3; Agere, 2; Raytheon, 2; Huawei, Hitachi, Motorola), to RF and microwave products manufacturers (RM Microdevices, 6 patents, Andrew LLC, 7 patents, Atheros, 2 patents, Scientific Atlantic).
- Image and Video Processing Patents: 5,689,591, 5,854,856, and 6,760,488. These patents disclose fundamental work in image and video representations and processing. Images and video are commonly treated as large sets of pixels. These pixel based representations tax resources: transmission of images and video requires large bandwidth; to interpret them necessitates too much computing; or to find in an image or video a relevant object may take too long. These patents disclose generative video (GV), a content and context based representation for images and video. Generative video represents images and videos by objects (for example, background and foreground, moving objects, objects at different depths); then replaces the video sequence by the panoramic background and the objects extracted from the video sequence. These patents disclose successful methods to extract the GV representation of a video by stitching the background across images in the sequence, by segmenting objects from the background, and by estimating the motions of the objects. To achieve 3D object representations, patent 6,760,488 discloses a very efficient, high quality method to extract 3D objects from monocular 2D video sequences. Background/ objects representations are now common in image standards like MPEG4 and MPEG7. These video patents (5,689,591; 5,854,856; 6,760,488) have been referenced as prior art by at least 46 patents from over 20 companies in the video, networking, internet spaces, including Google (4), Philips (4) Matsubishi (3), Microsoft (3), Sharp (2), IBM (1), Cisco (1), HP (2), Honeywell (2), Samsung, among others.
- Magnetic Resonace Imaging (postprocessing) patents: 7,734,075 and 7,394,921. These patents develop post-processing automatic methods for segmentation of human organs like liver and kidneys in Magnetic Resonance Imaging. These patents are owned by Siemens and the technology was developed for their MRI machines.
Patents issued or filed with the US Patent Office:
- "Collaborative Tracking in Cluttered Indoor Environemnts," Usman Khan, Soummya, Kar, José M. F. Moura, provisional patent filed with the US Patent Office, November 9, 2009; regular application filed November 2010.
- “Time Reversal Beamforming For Cancer Detection Using Electromagnetic Radiation,” Yuanwei Jin and José M. F. Moura, filed with the US Patent Office, July 9, 2008, S.N. 12/217,838.
- “Application of Time Reversal to Synthetic Aperture Radar,” Yuanwei Jin and José M. F. Moura, Patent and Trademark Office, S.N. 7,928,896, issued April 19, 2011; initially filed July 9, 2008, S.N. 12/217,839.
- “Contrast-Invariant Registration of Cardiac and Renal Magnetic Resonance Perfusion Images,” Ying Sun, Marie-Pierre Joly, and José M. F. Moura. Patent and Trademark Office, S.N. 7,734,075, issued June 8, 2010; filed originally with US Patent and Trademark Office, US Serial No. 11/078,035, March 11/2005.
- “Integrated Registration of Dynamic Renal Perfusion MR Images,” Ying Sun, Marie-Pierre Joly, and José M. F. Moura, Patent and Trademark Office, S.N. 7,394,921, issued July 1, 2008; foreign patent filed in The Netherlands, March 2004.
- "System and Method for Generating a Three-dimensional Model from a two-dimensional
Image Sequence," Pedro M. Q. Aguiar and Josť M. F. Moura,
US Patent and Trademark Office, S.N. 6,760,488, July 6, 2004.
- "Soft and Hard Sequence Detection in ISI Memory Channels," José
M. F. Moura and Alek Kavcic, US Patent and Trademark Office, S.N. 6,438,180,
August 20, 2002.
- "Method and Apparatus for Correlation-Sensitive Adaptive Sequence
Detection," Aleksandar Kavcic and José M. F. Moura, US Patent
and Trademark Office, S.N. 6,201,839, issued March 13, 2001.
- "Adaptive Parametric Signal Predistorter for Compensation of Time
Varying Linear and Nonlinear Amplifier Distortion," José M.
F. Moura, John Stonick, and Virginia Stonick, U.S. Patent and Trademark
Office, S.N. 5,900,778, issued May 4, 1999.
- "Generative Video: Very Low Bit Rate Video Compression," Josť M. F. Moura
and Radu S. Jasinschi, US Patent and Trademark Office, S.N. 5,854,856, issued
December 29, 1998.
- "Noncausal Predictive Image Codec," Nikhil Balram and José
M. F. Moura, US Patent and Trademark Office, S.N. 5,689,591, issued November
- "An Efficient Implementation of a Video Codec Compliant to H.263
and H.263 Version 2," Deepak Turaga, Tsuhan Chen, and José M.
F. Moura, disclosure filed with the CMU Technology Transfer Office, January
- "Robust Receivers," Chuang (Mike) He and José M. F. Moura,
Invention disclosure filed with the CMU Technology Transfer Office, December