PTAB
IPR2017-01190
Samsung Electronics Co Ltd v. Image Processing Technologies LLC
Key Events
Petition
Table of Contents
petition
1. Case Identification
- Case #: IPR2017-01190
- Patent #: 6,717,518
- Filed: March 29, 2017
- Petitioner(s): Samsung Electronics Co., Ltd.; and Samsung Electronics America, Inc.
- Patent Owner(s): Image Processing Technologies, LLC
- Challenged Claims: 39
2. Patent Overview
- Title: Process and Apparatus for Detecting the Drowsiness of a Person
- Brief Description: The ’518 patent discloses a process for detecting driver drowsiness by analyzing video images of the driver's face. The system identifies a facial characteristic, uses an anthropomorphic model to define a sub-area containing the eyes, and analyzes histograms of pixels within that sub-area over time to detect eye features and blink rates.
3. Grounds for Unpatentability
Ground 1: Obviousness over Eriksson and Stringa - Claim 39 is obvious over Eriksson in view of Stringa.
- Prior Art Relied Upon:
- Core Argument for this Ground:
- Prior Art Mapping: Petitioner argued that Eriksson, a system for detecting driver fatigue, disclosed most elements of claim 39. Eriksson taught acquiring an image, using a symmetry histogram to find the face's center, narrowing the search space to an "eye-region," selecting pixels based on intensity valleys, forming a horizontal intensity histogram, and analyzing it over time to detect a pupil and infer blinks. Stringa, which taught an eye localization algorithm for face recognition, supplied the missing or enhanced "anthropomorphic model" element by disclosing the use of a priori anthropometric data (e.g., relative positions of eyes and nose) to define "expectation zones" for the eyes.
- Motivation to Combine: A POSITA would combine these references because they addressed similar problems of eye localization in images. Petitioner asserted three primary motivations: (1) Eriksson's system had acknowledged weaknesses, such as "false alarms" when a driver's head was turned, which Stringa’s normalization procedures were designed to correct; (2) incorporating Stringa's more extensive use of anthropomorphic models would improve the precision of Eriksson's pupil localization; and (3) Eriksson explicitly cited Stringa, directly leading a POSITA to consider its teachings as a relevant and helpful reference.
- Expectation of Success: A POSITA would have expected success because the combination involved applying Stringa's known anthropomorphic modeling techniques to Eriksson's similar drowsy driving detection system to achieve the predictable result of improved eye localization accuracy.
Ground 2: Obviousness over Ando and Suenaga - Claim 39 is obvious over Ando in view of Suenaga.
- Prior Art Relied Upon:
- Core Argument for this Ground:
- Prior Art Mapping: Petitioner contended that Ando, a system for recognizing a driver's eyes to control vehicle devices, disclosed all limitations of claim 39. Ando taught identifying the head and forehead boundaries, using an anthropomorphic model based on those boundaries to define an eye search region, selecting pixels in that region, forming a histogram, and analyzing it over time to detect the pupil and eye state. Suenaga, which disclosed a system for detecting driver drowsiness, provided an alternative or supplementary teaching of using both X- and Y-histograms to determine the eye's open or closed state, improving on Ando's methods.
- Motivation to Combine: A POSITA would combine Ando and Suenaga as they were directed to the same problem of determining a driver's eye state from video. Ando's system was described as imperfect, particularly in "nonuniform illumination," motivating a POSITA to seek improvements. Suenaga provided such an improvement by using X- and Y-histograms to increase detection accuracy. Petitioner also noted a patent citation link: Suenaga cited Patent 5,293,427 ("Ueno"), which in turn cited Ando, suggesting a POSITA researching the field would have readily discovered both references.
- Expectation of Success: The combination would predictably yield success because it involved applying Suenaga's known X and Y histogram techniques to Ando's pupil detection system, retaining the original functions while simply providing an additional, known technique to increase detection accuracy.
Ground 3: Obviousness over Ando and Stringa - Claim 39 is obvious over Ando in view of Stringa.
- Prior Art Relied Upon:
- Ando (Patent 5,008,946)
- Stringa (Luigi Stringa, Eyes Detection For Face Recognition, Applied Artificial Intelligence, 1993)
- Core Argument for this Ground:
- Prior Art Mapping: Similar to Ground 2, Petitioner argued that Ando disclosed all elements of claim 39. The combination with Stringa was presented to further strengthen the obviousness case. Stringa's advanced techniques for using horizontal intensity histograms and its explicit use of a priori anthropometric information to define "expectation zones" for the eyes would be combined with Ando's foundational system for tracking eye state over time.
- Motivation to Combine: A POSITA would combine Ando and Stringa to improve Ando's admittedly imperfect face detection system. Ando's primary focus was on developing algorithms, and a POSITA would have sought later improvements like those in Stringa. Stringa's pupil detection technique, using horizontal intensity histograms, was argued to be more advanced than Ando's. A POSITA would have been motivated to incorporate Stringa's more robust method to increase the pupil detection accuracy and reliability of Ando's system.
- Expectation of Success: A POSITA would have expected the combination to work predictably, as it involved applying known techniques (Stringa's horizontal pupil histograms) to a similar system (Ando's eye state detector). The combined elements would retain their original functions, with Stringa simply providing a more advanced technique to enhance detection accuracy.
4. Relief Requested
- Petitioner requests institution of an inter partes review and cancellation of claim 39 of the ’518 patent as unpatentable.
Analysis metadata