1994 U.S. Software Patent Statistics

by Gregory Aharonian

The Internet Patent News Service has completed its analysis of U.S. software patents issued in 1994. The final total was 4569, pushing the total number of software patents issued since 1970 to over 15,000. As the following list shows, hot topics for software patenting include object-oriented programming, interactive television, networking, health care, and automobiles. The year's software patents were issued in the following categories (some patents were counted in more than one category):

Patents Issued

Category

23 Image processing
32 Networks/communications
48 Operating systems
74 Process/numerical control
37 Graphics
92 Medical/health care
41 Engineering
32 Automobiles/transportation
23 Graphical user interfaces
11 Signal processing
73 Database
66 CASE
62 Security/encryption
49 CAE/circuit design
47 Financial/management
51 Office automation
29 CAD
21 Word processing
19 Physics
3 Navigation
95 Speech recognition/synthesis
92 Robotics
89 Neural networks
88 Distributed processing
86 Pattern recognition
78 Compression
78 Artificial intelligence
71 Biology
70 Music
67 Natural-language analysis
66 Numerical analysis
62 Character recognition
57 Multiprocessing
56 Algorithms
51 Chemistry
49 Object-oriented programming
48 Games
47 Geophysical
47 Fuzzy logic
38 Simulation
34 Vision
33 Education
28 Parallel programming
26 Virtual reality
8 Spreadsheets
6 Biotechnology

Corporate software-patent assignees (* indicates Japanese):

Patents Issued

Corporation

96 IBM
89 *Hitachi
7 DEC
7 Xerox & Fuji Xerox
7 *Toshiba
97 Hewlett-Packard
82 *Fujitsu
70 *Canon Kabushiki Kaisha
68 Motorola
68 *Matsushita Electric Industrial
65 *Mitsubishi Denki Kabushiki
61 ATT Bell Laboratories
50 General Electric
49 *Ricoh
39 Eastman Kodak
39 *Sharp Kabushiki Kaisha
36 U.S. Navy/Army/Air Force
36 *Fanuc
35 Sun Microsystems
34 Intel
32 *NEC
30 *Yamaha
30 *Honda
27 *Brother Kyogo Kabushiki Kaisha
27 Texas Instruments
24 Bell Communications Research
23 Sony
22 Hughes Aircraft
21 Microsoft
21 Ford Motor
20 Siemens Aktiengesellschaft
20 Samsung Electronics
20 Apple Computer
19 Honeywell
17 *Fuji Photo Film
17 *Casio Computer
16 Westinghouse Electric
16 Unisys
15 Schlumberger Technology
14 *Zexel
14 *Pioneer Electronic
13 E.I. Du Pont de Nemours
13 VLSI Technology
13 Pitney Bowes
13 *Dainippon Screen Manufacturing
13 *Konica
13 *Toyota Jidosha Kabushiki Kaisha
13 Rockwell International
12 Compaq Computer
12 *Nissan Motor
12 France Telecom

The flood continues in 1995, with an all-time high of 125 software patents issued in the first week of January. My prediction is that 5500 software patents will be granted in 1995, giving a two-year total of 10,000 software patents --- and this in an industry where the last original software idea was published ten years ago.

Unfortunately, the Patent and Trademark Office (PTO) is making little progress on the complex problem of software prior art. The average number of nonpatent prior-art items cited in U.S. software patents rose from 1.8 to 2.0 items per patent in 1994; still a tiny fraction of the 50 or more nonpatent prior-art items per patent that an experienced prior-art researcher can locate. Analysis of the 1994 software patents using the largest software prior-art database in the country (maintained by Source Translation & Optimization, publisher of the Internet Patent News Service) shows that over 70 percent of the 1994 software patents should have had one or more of their claims disallowed on prior-art grounds. As such, expect more and more software infringement lawsuits in the years to come.

To illustrate, I have selected two of the worst software patents recently issued. The claims and abstracts are reflective of the patent as a whole. They also typify the large group of software patents issued that cite, on the average, two nonpatent prior-art references where two dozen should be cited.

Source Translation & Optimization has just released a report detailing over a dozen statistical views of the problems of prior art for U.S. software patents. The report, titled ``Demographics of US Software Patents: Foundation for Software Prior Art Planning,'' is available from STO, P.O. Box 404, Belmont, MA 02178, for $100.00.

Patent #5,365,473

Integral value calculating device and function gradient calculating device.: Fujitsu Limited, Kawasaki, Japan; Filing date: November 27, 1992; 6 prior patent references; 0 prior non-patent references; PTO group 2360

What is claimed:

1. An integral value calculating device for calculating an integral value of a function, comprising:

a computer, comprising:
graph display means for outputting a graph for a function according to received data;
calculation range input means, communicating with said graph display means, for designating an integral range for said function according to position data externally designated for said graph, and;
integral value calculating means, communicating with said calculation range input means, for calculating the integral value of said function for said designated integral range, and;
an external unit connected to said computer and receiving the graph and the integral range.

Abstract: An integral value calculating device is used for calculating an integral value of a function, and a function gradient calculating device is used for calculating the gradient of the graph for a function. The object is to obtain a transport coefficient, etc. by designating an appropriate calculation range in analyzing the data output by a molecular dynamics simulator. In the integral value calculating device, a graph display unit outputs a graph according to received data, for example, a graph of a time correlation function F(t) of the time differential W'(t) to the dynamics volume W(T), in the format applicable to an external unit, for example, a display unit. A calculation range input unit designates an integral range for the function according to the position data externally input for the graph. An integral value calculating unit calculates an integral value for a function F(t), for example, for the designated integral range. In the function gradient calculating device, two points on a graph in which the gradient should be calculated are designated according to the position data input for a graph displayed on the display unit, for example. Then, a function gradient calculating unit calculates the gradient of the graph.

In this patent, filed in 1992, the first claim basically describes dozens of commercially available mathematical-analysis programs released in the 1980s --- MathCad, Macsyma, MathLab, Gauss, IMSL, and others --- programs that display an equation on the screen, allow the user to choose two endpoints, and then perform an integration and display the results. Engineering journals have published thousands of papers over the past 20 years that directly relate to this claim. For example, the papers ``Computer Graphics Assisted Numerical Analysis Instruction'' (Computers & Graphics, 1976) and ``Interactive Computing and Graphics in the Interpretation of Geomagnetic Spectra'' (Physics of Earth and Planetary Interiors, 1976) relate to the first claim, and its use for physics data analysis. The technique described in this patent is also described in several journals on numerical analysis, several physics journals dealing with quantum chemistry and molecular dynamics, numerous books (including Numerical Recipes, which the PTO has cited in a few other patents), and conference proceedings.

Still, neither Fujitsu, one of the largest computer companies in the world, nor the PTO cited any of these materials as prior art, and probably did not even bother to look. This first claim is too obvious: It basically patents graphical numeric integration. Even if the rest of the claims and specification restrict its scope to molecular dynamics simulation, there is still prior art (see Computer Physics Communications, or the QCPE library).

Patent #5,374,996

Image Processing System: Matsushita Electric Industrial, Osaka, Japan; Filing date: June 26, 1991 (in Japan); 12 prior patent references; 0 prior non-patent references; PTO group 2170

What is claimed:

1. An image-processing system, comprising:

reading means for outputting multi-level image data corresponding to a density of each of pixels, to which an image of a draft is partitioned;
sampling means for obtaining density data by sampling the image data output from the reading means at a predetermined rate of the number of the pixels, which correspond to image data to be sampled, to that of all of the pixels;
storing means for storing the density data obtained by the sampling means;
image processing means for performing a processing on the image data output from the reading means;
displaying means for displaying an image represented by data obtained as a result of the processing by the image processing means;
indicating means for indicating a position in the image displayed by the displaying means; and

control means for reading the density data corresponding to the position indicated by the indicating means from the storing means and making the displaying means display the read density data.

Abstract: An image processing system for reading multi-level image data corresponding to each pixel of an input image from a multi-tone-level draft image such as a photograph and correcting a density level at each of the pixels, which is represented by the obtained image data, and outputting data representing corrected density levels. The image-processing system first performs an appropriate sampling of the read image data of the pixels at a sampling rate in such a manner to enable reference to the density levels at the pixels when correcting the density levels. The sampling rate is most appropriately predetermined.

The first claim in this patent includes nearly every image-processing system described since the 1970s, including the 1000+ image-processing patents and 5000+ image-processing journal articles and technical reports. Citing just 12 prior patents and no nonpatent prior-art references --- given the broad, obvious structure of claim 1 --- is a mistake. Maybe somewhere in the dependent claims there is something novel, but if so, it should be brought up into the independent claim. The PTO should have a blanket policy of rejecting every new image-processing patent it gets --- there is just nothing left to invent (applications included) that hasn't been previously published.

Gregory is editor of the Internet Patent News Service provided by Source Translation & Optimization. Gregory can be contacted at patents@world.std.com.