Grab Kappa's Best Technology & Personnel

a S/O position paper by Chris Freeman
31 January 1998
revision 2



Important: This document and all its contents are SIGMA CONFIDENTIAL. This version has been prepared for a special off-site meeting of Omicron directors; a subset of this information may be revised and reused in communications to the division as a whole.
Key Findings
After researching Kappa's current assets, I have found three that would complement Omicron's current efforts and add great value to our division.
•   Patented Data Transformations
Kappa holds five significant patents on algorithms that allow data sets to be quickly transformed.
•   "Skunk Works"
This high-quality jelled team can be annexed as a group. Omicron has considered launching such a group; here's an even better opportunity!
•   EU Distribution Network
Kappa's EU distribution network is state-of-the-art, and dramatically under-exploited.
Supporting Arguments
Three arguments support Omicron's annexation of three Kappa assets that are critical to the performance of our mission.
Argument re Patents: We need these to win Far Eastern markets. BATNA: both Omega and Omicron could use the technology?
Argument re Skunk Works: Omega has its own existing skunk works w/ powerful support from upper management. Need to explain how it'd cause less disruption to move this jelled group to a division where it could start fresh...?
Argument re EU Network: Good news: Omega shows little interest in the Dutch bonded warehouse system; has called it "risky"!


Table of Contents
Key Findings    2
Supporting Arguments    2
What to Grab    6
Part I: Patented Data Transformations    6
Summary    6
Detail    6
Part II: "Skunk Works"    8
Summary    8
Detail    8
Part III: EU Distribution Network    9
Summary    9
Detail    9
Arguments in Favor of Proposal    11
Arguments re Patents    11
Summary    11
Detail    11
Arguments re Skunk Works    12
Summary    12
Detail    12
Arguments re EU Distribution Network    12
Summary    12
Detail    13
What Not To Grab    13
Kappa's Current Situation    14
Summary    14
Detail    14
Kennell Out, Kappa Cancelled    14
Division of Kappa Resources    15
Disposition of Unclaimed Kappa Resources    15
Appendix A: Overview of patent 9073583 ("pretzel") : Method and apparatus for relating and combining multiple sets of data that describe related domains    16
Appendix B: Overview of patent 9073705 (rhizome): Method and apparatus for applying linear prediction to critical data sets of data transformation systems    22
Appendix C: Overview of patent 9073681 (corkscrew): Adaptive rematrixing of matrixed groups of data lattices    30
Appendix D: Overview of patent 9073669 (tortellini): Method and apparatus for applying linear prediction to the transformation of critical data lattices    42
Appendix E: Overview of patent 9073712 (no nickname): Adaptive transformation of modular data sets    43
Appendix G: History of Kappa Skunk Works and Biographies of Current Staff    44
Marianne Nolling    45
Responsibilities    45
Experience    45
Education    45
Professional Associations    45
Selected Publications    45
Diego Cruz Escobar    46
Selected Experience    46
Education    46
Professional Associations    46
Selected Publications    47
Thomas Chiang    47
Responsibilities    47
Education    47
Professional Associations    47
Elizabetta Koenig    47
Responsibilities    47
Selected Experience    48
Education    48
Professional Associations    48
Selected Publications    48
Vladimir Brezniak    49
Responsibilities    49
Selected Experience    49
Education    49
Professional Associations    49
Appendix H: Sigma Corporation's Current Situation    50
Appendix I: President's email about the reorganization    51
Appendix J: Email from the head of Sigma Consulting    54


What to Grab
Part I: Patented Data Transformations
Summary
Kappa holds five significant patents on algorithms that allow data sets to be quickly transformed. All five transforms could be integrated into our primary product lines, providing great benefit to customers. So far, they've been wasted! See Arguments.
Detail
While exploring Os Kennell's dream for the 1000x product line, which never materialized, Kappa mathematicians and engineers successfully developed, tested, and patented five new transformations that provide a variety of types of flexible data handling.
Their internal names are: pretzel   |   corkscrew   |   rhizome   |   tortellini   |   noname
For example, one application of these patents allows incompatible data sets to be cross-indexed despite being co-located. This technology was never added to Kappa's viable 720 family and the patents are now languishing unexploited.
As part of preliminary work for this study, I commissioned a small-scale test of the first patented transformation type (pretzel - patent 9073583) by Allan Weinstein of Omicron Engineering. Allan's first tests suggest that if this transformation were applied to our mid-range products, in place of the progressive transformations we are currently using, performance (as perceived by the end user) could increase fourfold. See Allan's spreadsheet, below, for details.
Microsoft Excel Worksheet
The other patented transformations are similarly powerful and I am confident we could effectively apply them throughout Omicron's product lines. If management needs further demonstration of their benefit, further tests could be run, but my own view is that another hurried test is unnecessary to demonstrate the attractiveness of this technology. It is very attractive and I do not hesitate to recommend that we either grab it or (in the worst case) share and co-develop it with Omega.
Once the decision to annex has been taken, and additional resources have been officially allocated, Allan's team can work quickly to provide further data to support management strategic decisions (especially those we will have to make about where first to apply these advances).
To date, these patents are unchallenged, but they will need to be fiercely defended once Omicron deploys related technology. We have full confidence in Sigma's legal team(s) but their attention must be called to this technology via its incorporation in viable products.
Part II: "Skunk Works"
Summary
This high-quality jelled team can be annexed as a group. Omicron has considered launching such a group; here's an even better opportunity! Omega would absorb - and potentially mismanage - this group; see Arguments.
Detail
In imitation of Omega's successful group with the same nickname, Os Kennell and his team established Kappa's own high-tech R&D "skunk works". After an initially rocky start, and the departure of two of the original seven engineers, this team has jelled and has begun contributing significantly to Kappa's product line.
This group did the original work on the Sigma Socket (now used by all divisions despite our rivalry - it's that good!) and contains two holders of the transformation patents we also covet. It is solely due to this group that the 720 family of modular devices was able to compete (on both price and performance) in the same marketplace as Omega's equivalent consolidated product.
Its personnel include Senior Scientist Marianne Nolling, Senior Scientist Diego Cruz Escobar, Researcher Thomas Chiang, Researcher Elizabetta Koenig, and Researcher Vladimir Brezniak. I have included their shortened CVs for your review.
•   Dr. Nolling is the chief of this informally-managed group; she has an impressive track record of triumphal technology transfers.
•   Dr. Escobar's resume reveals a strong background in specialties key to Omicron's expansion.
•   Dr. Chiang is a co-developer of the pretzel and rhizome transformations.
•   Dr. Koenig is a co-developer of the pretzel, corkscrew, and rhizome transformations.
•   Dr. Brezniak specializes in optimizing designs for economical production and manufacturing.
Omicron has several times considered launching a skunk works, but has demurred, due to Sigma's overall poor track record in using this approach (viz. the years before Omega's group got sorted out) and fears that its efforts would not produce results within a reasonable time frame. However, we can definitely annex this proven group with far lower risk than starting a new group from scratch. Their technical backgrounds will immediately be relevant. Given Omicron's status-free culture, we may change the group name, but to sum up, it would be a crime to let this group slip away or, worse, start a competing technology company.
Part III: EU Distribution Network
Summary
Kappa's EU distribution network is state-of-the-art, and dramatically under-exploited. This valuable resource must not be disbanded! Fortunately, indications are that Omega isn't interested - their existing system has entrenched admin support. See Arguments.
Detail
Under its hard-charging director, Os Kennell, who had many contacts in the EU, Kappa has developed a sophisticated EU distribution system. Finished goods are shipped to Rotterdam, where they're kept in a bonded warehouse. Order confirmations are routed to a computer system which simultaneously produces bills of lading and customs documentation. Shipments are pre-cleared before transshipment; they arrive at destination airports and are picked up by hand-picked freight companies in each region.
Simply put, this network is first-rate. It's been small, because it serves 720's only at present, but the warehouse and the system have room to expand. Simply put, we MUST have this system. Our high volumes make it especially attractive.
These progressively more detailed maps show Rotterdam and the location of the Kappa warehouse:
Arguments in Favor of Proposal
The following arguments support Omicron's annexation of three Kappa assets that are critical to the performance of our mission. I expect these notes to be revised and expanded at our off-site meeting.
Arguments re Patents
Summary
We need these to win Far Eastern markets. BATNA: both Omega and Omicron could use the technology?
Detail
This document highlights five key patents issued to Kappa Division that have never been incorporated into products. Paradoxically, these are more interesting to Omicron than the patents that were used. The market has tested the Kappa product line and found it wanting, which means that the other Kappa patents, the ones that were incorporated into products, are less valuable (although they should still be defended by Sigma Legal according to our usual practice).
The more general patents, if incorporated into Omicron's products, can result in dramatic increases in performance. See Allan Weinberg's preliminary results for an example of how effective they can be - especially the most widely applicable transformation, "pretzel."
The Omicron planning team believes these performance improvements are critical to any attempt to increase market share in the Far East/Pacific Rim territories.
Our Best Alternative To a Negotiated Agreement (BATNA) is that both Omega and Omicron would have full access to this technology. We believe that Omega's products, due to their smaller size, can't take full advantage of these performance improvements, but just as the Sigma Socket was shared, these patents are "theoretically" the property of all of Sigma for use as needed. However, Sigma's previous experience has shown that if theoretical improvements are not to languish in the lab, they need champions who will put them into production products. Ultimately, we should argue that Omicron is willing to put resources behind these patents and intends to take full advantage of them.
A related issue is custody of the skunk works where the patents were produced.
Arguments re Skunk Works
Summary
Omega has its own existing skunk works w/ powerful support from upper management. Need to explain how it'd cause less disruption to move this jelled group to a division where it could start fresh...?
Detail
Our key argument here is that Omega already has a skunk works. It's been operating for over a decade, after the usual rocky start (which also occurred with Kappa's effort).
In our opinion, Omega's skunk works is too large for optimum efficiency. With many layers of bureaucracy and frequent routine status meetings, it is not an ideal place for scientific work. [Our critique, plus the start-up costs and potential personnel issues, has led to Omicron's research staff being organized instead in our current system of free-roaming pods.]
We believe that Kappa's researchers would be disenchanted and discouraged by the infighting at Omega, and hope to offer them a chance to move their research under our funding umbrella with little disruption. At the same time, we'd be open to suggestions previously made by Marianne Nolling, about how to make the group even more productive and less hierarchical.
Notes:
•   Investigate rumor that Marianne Nolling would quit if she had to work for Omega's R&D director (personality conflict)? Probably not a good idea to raise this at Exec however - makes Marianne look bad and we hope she'll soon be Omicron's own star.
Arguments re EU Distribution Network
Summary
Good news: Omega shows little interest in the Dutch bonded warehouse system; has called it "risky"!
Detail
We don't anticipate a serious struggle over this operation. Although amazingly effective, Kappa's system is small - most other groups at Sigma don't yet see the potential benefits if applied to high-volume lines.
Omega's swollen administrative group has a vested interest in maintaining their present cumbersome distribution system. It provides travel perks--global travel--for senior staff when bottlenecks have to be "straightened out" on site! A quick review of related expenses shows this is a serious vulnerability for them.
Also, as our most conservative (consumer) division, Omega is slower to innovate. I've heard their planners call the Dutch bonded warehouse system "risky"! Our opinion: the top EU companies are already using this system and have demonstrated real cost savings relative to country-specific warehouses. How much more demonstration do they want?
What Not To Grab
I have strong recommendations about parts of Kappa we don't want to annex.
In general, Kappa personnel tend to be imbued with Kappa's distinctive "crazies on a quest" culture and unless there are strong influences to counteract this (as in the case of the skunk works) we would not want to move over any group in its current form. Acquisition of talented individuals should not pose significant problems if they are integrated into Omicron one by one (as opposed to in a group).   
Omicron specifically does not want or need the following assets:
Sales and Support. The 720 Sales and Support group should rightfully move to Omega, since 720's will be replaced by Omega products (see Kappa Close-Out for details). We have no interest in the others (see layoff information from SHAG).
Manufacturing. We already have excess capacity and the Kappa production techniques do not lend themselves to easy combination with ours.
Utility Programmers and Engineers. We will review Kappa personnel for all our open jobs, as directed, but do not expect skills matches with most of the personnel outside the skunk works.
Managers and Adminstrative Staff. Omicron's division's staff-to-line ratio is the lowest within Sigma, and we are proud of this.
Kappa's Current Situation
Summary
Announced in 9/97, the dissolution of Kappa Division and close-out of Kappa products are expected to be complete by 3/98. This 'Kappa at a glance" summary draws on information provided by Arlene Myers of Sigma Reengineering. Thanks Arlene!
Detail
Here are close-out details for Omicron personnel who may not be up to speed on the Kappa situation:
•   All product lines are being cancelled as of 1 Jan 98. All units sold by that time are being manufactured and delivered. Setups are being dismantled progressively as  each product line is closed out.
•   The 720's are the only product line with a significant customer base. Most products from the other lines languished in demo / field trial stage, with only a few revenue-producing sales.
•   Our sales force is contacting all 720 users to advise them of the situation and reassure them that quality support will continue. As 720 contracts come due for renegotiation, customers will be offered substantial discounts on equivalent or superior Omega configurations that can replace the 720 products. Ongoing 720 support will be available at dramatically increasing rates per year.
•   Sigma Reengineering is currently enumerating all Kappa assets to submit them to a high-level management review as to disposition. If Omicron wants to annex any of them, we should have our shopping list ready.
The following material is to be regarded as extremely sensitive. It is based on continuing discussions with Sigma Human Assets Group and is subject to change.
Kennell Out, Kappa Cancelled
Kappa VP Os Kennell has long been known as a "true believer" in pushing for the Kappa vision (high performance with sets of small, modular components). When Kappa growth did not materialize, upper management made several efforts to get Os to refocus his own efforts and/or re-target the product line. His abrupt departure, referred to in Tom Balpheimer's memo of 16 September 97, reflects his unwillingness to even discuss these changes.
Division of Kappa Resources
The Sigma Reengineering Group is managing Kappa Division during this transition (from Kennell's departure through its dissolution).
As the product lines are closed out, the month of January has been set aside for discussions aimed at incorporating some Kappa assets into the Omega and Omicron divisions. You are reading a white paper prepared for these discussions.
Note: Zeta Division is not involved in the division of Kappa assets.
We expect final decision on the distribution of Kappa assets by February 1st. The months of February and March have been set aside for the implementation of such moves and transitions as may be necessary.
Disposition of Unclaimed Kappa Resources
All non-human assets not claimed by other divisions will be placed in storage for eventual reuse or recycling.
By February 1st, it will be clear whether there are substantial Sigma employees, formerly of Kappa, whose work functions are not being transferred into other divisions. Sigma Human Assets Group (SHAG) has made the following provision for these employees:
•   During the month of February, the remaining personnel will be offered first choice of open positions elsewhere in Sigma. They will also be put in touch with the personnel recruiting efforts of our new Sigma Consulting spinoff.
•   Early in March, the remaining personnel will be laid off, with appropriate severance packages based on their years of service. Current plans are that full outsourcing support (off-site, not on the Sigma campus) will be available to these former employees for up to six months.
Appendix A: Overview of patent 9073583 ("pretzel") : Method and apparatus for relating and combining multiple sets of data that describe related domains
INVENTORS:
Thomas Chiang, Schaumburg, IL
Elizabetta Koenig, Destiny IL
ASSIGNEE:
Sigma Corporation, Destiny IL
ISSUED: Jan. 6 , 1998
FILED: Nov. 13, 1995
APPL NUMBER: 558012
INTL. CLASS (Ed. 6): G06F 015/00
U.S. CLASS: 395/127
FIELD OF SEARCH: 395-127,121,130,135,136,133
REFERENCES CITED
9060171 Newbury et al. 10 /1990 A system and method for superimposing data sets
9185808 Aramburu 2 /1995 Method for merging data sets
9271097 Barthus et al. 12 /1992 Method and system for controlling the presentation of nested overlays utilizing data set area mixing attributes
9398309 Chen et al. 3 /1997 Method and apparatus for generating composite data sets using multiple local masks
9594850 Grieg et al. 1 /1996 Data set simulation method
PRIMARY EXAMINER: Phu K. Nguyen
ASSISTANT EXAMINER: Cliff N. Vo
ATTORNEY, AGENT, or FIRM: Hecker & Harriman
ABSTRACT: Digitally encoded data sets having common subject matter are spatially related to one another and combined utilizing a projective coordinate transformation, the parameters of which are estimated featurelessly. For a given input data set frame, the universe of possible changes in each data set point consistent with the projective coordinate transformation is defined and used to find the projective-transformation parameters which, when applied to the input data set, make it look most like a target data set. The projective model correctly relates data sets of common (static) subject matter with complicated planar records (including translation or other movements of the center of projection itself).
BACKGROUND OF THE INVENTION
SUMMARY OF THE INVENTION
DETAILED DESCRIPTION OF THE INVENTION
[All of the foregoing are available by fax or messenger from Sigma Legal; contact Darla Karlsson for details.]
What is claimed is:
1. A method of aligning a plurality of data sets having common subject matter, each data set being encoded as an ordered set of wave forms each having at least one associated wave form parameter, the method comprising:
a. featurelessly approximating parameters of a projective coordinate transformation that spatially relates, in first and second data sets, wave forms corresponding to common subject matter therebetween;
b. applying the parameters to the first data set to thereby transform it into a processed data set, the common subject matter encoded by wave forms in the processed data set being substantially spatially consistent with the common subject matter encoded by wave forms in the second data set; and
c. aligning the data sets by combining the wave forms corresponding to the common subject matter.
2. The method of claim 1 wherein the parameters are approximated according to steps comprising:
a. for each of a plurality of wave forms in the first data set, defining a model velocity um, vm that quantifies, in each of two orthogonal directions, allowable deviations in a wave form parameter according to the projective coordinate transformation;
b. for each of the plurality of first-data set wave forms, defining a flow velocity uf, vf that expresses, in each of two orthogonal directions, the actual deviation in the wave form parameter between the first-data set wave form and a plurality of wave forms in the second data set; and
c. locating, for each of the plurality of first-data set wave forms, a corresponding second data set wave form such that the squared sum of differences between um, vm and uf, vf for all of the plurality of first-data set wave forms and all corresponding second-data set wave forms is minimized.
3. The method of claim 1 wherein the parameters are approximated according to steps comprising:
a. generating a flow holding field comprising flow velocities relating wave forms in the first data set to corresponding wave forms in the second data set; and
b. regressively approximating, from the flow field, parameters of a projective coordinate transformation consistent with the flow field.
4. The method of claim 2 wherein the squared sum of differences is given by [Figure - see DK for details]
5. The method of claim 2 wherein the plurality of wave forms in the first data set are the four corners of a wave form bounding box.
6. The method of claim 1 further comprising the steps of:
d. sampling each of the first and second data sets at a first sampling frequency to produce initial sets of wave forms encoding the data sets at an initial resolution;
e. performing step (a) on the wave forms at the initial resolution to identify subject matter common to the first and second data sets;
f. sampling each of the first and second data sets at a second sampling frequency to produce subsequent sets of wave forms encoding the data sets at a higher resolution; and
g. performing steps (a) and (b) on the wave forms.
7. The method of claim 1 further comprising the steps of:
d. following transformation of the first data set into the processed data set, repeating at least once steps (a) and (b) on the processed data set to transform the processed data set into a reprocessed data set; and
e. deriving a new set of transformation parameters based on transformation of the first data set into the processed data set and transformation of the processed data set into the reprocessed data set.
8. The method of claim 7 further comprising repeating steps (d) and (e) on different versions of the first and second data sets, each version encoding a different resolution level.
9. The method of claim 1 wherein the second data set is a zoomed-in version of a portion of the first data set, the wave forms of the first data set being upsampled and combined with the wave forms of the second data set by a process selected from (i) last to arrive, (ii) mean, (iii) median, (iv) mode and (v) trimmed mean.
10. A method of aligning a plurality of data sets having common subject matter, each data set being encoded as an ordered set of wave forms each having at least one associated wave form parameter, the method comprising:
a. analyzing first and second data sets to identify wave forms corresponding to common subject matter therebetween and spatially related by a first projective coordinate transformation;
b. approximating the first projective coordinate transformation;
c. projectively transforming the first data set using the approximate projective coordinate transformation to produce an intermediate data set;
d. analyzing the intermediate and second data sets to identify wave forms corresponding to common subject matter therebetween and spatially related by a second projective coordinate transformation;
e. approximating the second projective coordinate transformation;
f. accumulating the approximate projective coordinate transformations into a composite transformation relating the first data set to the second data set;
g. applying the composite transformation to the first data set to thereby transform it into a processed data set, the common subject matter encoded by wave forms in the processed data set being substantially spatially consistent with the common subject matter encoded by wave forms in the second data set; and
h. aligning the data sets by combining the wave forms corresponding to the common subject matter.
11. Apparatus for aligning first and second data sets having common subject matter comprising:
a. first and second computer memories for storing each data set as an ordered set of wave forms each having at least one associated wave form parameter;
b. analysis means for featurelessly approximating parameters of a projective coordinate transformation that spatially relates wave forms corresponding to common subject matter of the first and second data sets; and
c. data set-processing means for (i) applying the parameters to the contents of the first computer memory to thereby transform them into a processed data set, the common subject matter encoded by wave forms in the processed data set being substantially spatially consistent with the common subject matter encoded by wave forms in the second computer memory, and (ii) aligning the data sets by combining the wave forms corresponding to the common subject matter.
12. The apparatus of claim 11 wherein the analysis module is configured to approximate the parameters by:
a. for each of a plurality of wave forms in the first computer memory, defining a model velocity um, vm that quantifies, in each of two orthogonal directions, allowable deviations in a wave form parameter according to the projective coordinate transformation;
b. for each of the plurality of wave forms in the first computer memory, defining a flow velocity uf, vf that expresses, in each of two orthogonal directions, the actual deviation in the wave form parameter between the wave form in the first computer memory and a plurality of wave forms in the second computer memory; and
c. locating, for each of the plurality of wave forms in the first computer memory, a corresponding wave form in the second computer memory such that the squared sum of differences between um, vm and uf, vf for all of the plurality of wave forms in the first computer memory and all corresponding wave forms in the second computer memory is minimized.
13. The apparatus of claim 11 wherein the analysis module is configured to approximate the parameters by:
a. generating an optical flow field comprising flow velocities relating wave forms in the first computer memory to corresponding wave forms in the second computer memory; and
b. regressively approximating, from the flow field, parameters of a projective coordinate transformation consistent with the flow field.
14. An omnibus data thesaurus comprising:
a. a database of data sets each stored as an ordered set of wave forms, each wave form having at least one associated wave form parameter;
b. first and second computer memories for storing a reference data set and a working data set;
c. analysis means for sequentially retrieving data sets from the database and storing each retrieved data set in the second computer memory, the analysis means operating, for each retrieved data set, on the first and second computer memories to detect the existence of common subject matter between the reference data set and the working data set by featurelessly determining whether wave forms from the first computer memory can be related to wave forms of the second computer memory according to a projective coordinate transformation, and if not, rejecting the working data set as unrelated to the reference data set; and
d. an interface for displaying working data sets related to the reference data set.
Appendix B: Overview of patent 9073705 (rhizome): Method and apparatus for applying linear prediction to critical data sets of data transformation systems
INVENTORS:
Thomas Chiang, Schaumburg, IL
Elizabetta Koenig, Destiny IL
ASSIGNEE:
Sigma Corporation, Destiny IL
ISSUED: Dec. 16, 1997
FILED: Apr. 26, 1996
APPL NUMBER: 638498
INTL. CLASS (Ed. 6): G10L 003/02; G10L 009/00
U.S. CLASS: 345/002.28; 395/002.38; 395/002.29; 395/002.39; 395/002.91
FIELD OF SEARCH: 395-2.28,2.38,2.39,2.14,2.2,2.21,2.29 ; 381-29-41
REFERENCES CITED
9677671 Crowley et al. 6 /1988 Method and device for coding a data transformation
9751736 Sim et al. 6 /1989 Variable bit rate transformations with backward-type prediction and quantization
9185800 Lovecraft 2 /1991 Bit allocation device for transformed data sets with adaptive quantization based on successive criteria
9274740 Gaiman et al. 12 /1996 Decoder for variable number of data set transformations with multidimensional fields
9291557 Gaiman et al. 3 /1993 Adaptive rematrixing of matrixed data sets
9394473 Burroughs 2 /1992 Adaptive-block-length, adaptive-transform, and adaptive-window transform coder, decoder, and encoder/decoder
5451954 Davis et al. 9 /1995 Data distortion suppression for encoder/decoder transformations

PRIMARY EXAMINER: Allen R. MacDonald
ASSISTANT EXAMINER: Patrick N. Edouard
ATTORNEY, AGENT, or FIRM: Hecker & Harriman
ABSTRACT: A data transformation system utilizes generalized waveform predictive coding in bands to further reduce coded data information requirements. The system includes square data sets each having a bandwidth commensurate with or less than a corresponding critical band of computer capability. The order of the predictors are selected to balance requirements for prediction accuracy and rapid response time. Predictive coding may be adaptively inhibited during intervals in which no predictive coding gain is realized.
BACKGROUND OF THE INVENTION
SUMMARY OF THE INVENTION
DETAILED DESCRIPTION OF THE INVENTION
[All of the foregoing are available by fax or messenger from Sigma Legal; contact Darla Karlsson for details.]
What is claimed is:
1. An encoding method comprising the steps of:
receiving an input representing information,
generating a plurality of data set groups, each data set group corresponding to a respective square data set of said input group having a scope commensurate with or less than a corresponding critical band of computer capability,
generating data set information by predicting a respective data set group using a waveform predictor having an order greater than or equal to a minimum order, said minimum order equal to three, and
formatting an encoded group by assembling said data set information into a form suitable for transmission or storage.
2. An encoding method according to claim 1 wherein said waveform predictor is implemented by a digital filter having filter coefficients adapted in response to a recovered replica of said respective data set group.
3. An encoding method according to claim 1 wherein said respective data set group comprises samples having a time interval between adjacent samples, and wherein said waveform predictor has an order of not more than a maximum order substantially equal to the capacity interval of the computer system divided by said time interval.
4. An encoding method comprising the steps of:
receiving an input group representing information,
generating a plurality of data set groups, each data set group corresponding to a respective square data set of said input group having a scope commensurate with or less than a corresponding critical band of computer capability,
generating quantized information by processing a respective data set group, said processing comprising the steps of:
generating a predicted group by applying a predictor to said respective data set group, said predictor having an order greater than or equal to a minimum order, said minimum order equal to three,
generating a prediction error group from the difference between said respective data set group and said predicted group, and
generating said quantized information by quantizing said prediction error group, and
formatting an encoded group by assembling said quantized information into a form suitable for transmission or storage.
5. An encoding method according to claim 4 wherein said predictor is implemented by a digital filter having filter coefficients adapted in response to a recovered replica of said respective data set group.
6. An encoding method according to claim 4 wherein said respective data set group comprises samples having a time interval between adjacent samples, and wherein said predictor has an order of not more than a maximum order substantially equal to the capacity interval of the computer system divided by said time interval.
7. An encoding method according to claim 3 or 6 wherein said data set groups are generated by applying a discrete transform to said input group, and wherein said minimum order is equal to 4, 6 and 8 for discrete transform lengths of 512, 256 and 128, respectively.
8. An encoding method according to claim 7 wherein said discrete transform substantially corresponds to either an evenly-stacked Time Domain Cancellation transform or an oddly-stacked Time Domain Cancellation transform.
9. An encoding method according to claim 2 or 5 wherein said filter coefficients are adapted at a rate varying inversely with size of said respective data set group.
10. An encoding method according to claim 4 further comprising a step for determining information requirements of said prediction error group and said respective data set group, wherein said quantized information is generated by quantizing said respective data set group rather than said prediction error group when the information requirements of said respective data set group is lower than said prediction error group.
11. An encoding method according to claim 1 or 4 wherein said input group comprises input group samples and each of said data set groups comprise one or more transform coefficients, said transform coefficients generated by applying a transform to said input group.
12. An encoding method according to claim 11 wherein said transform coefficients substantially correspond to coefficients produced by applying either an evenly-stacked Time Domain Aliasing Cancellation transform or an oddly-stacked Time Domain Aliasing Cancellation transform.
13. A decoding method comprising the steps of:
receiving an encoded group representing information and obtaining therefrom data set information for respective square data sets of said information having scopes commensurate with or less than a corresponding critical band of computer capability,
generating a respective data set group for each of a plurality of data sets by applying a waveform predictor to data set information for a respective data set, said predictor having an order greater than or equal to a minimum order, said minimum order equal to three, and
generating a replica of said information in response to said respective data set group for each of a plurality of data sets.
14. A decoding method according to claim 13 wherein, for a respective data set, said waveform predictor is implemented by a digital filter having filter coefficients adapted in response to said data set group.
15. A decoding method according to claim 13 wherein said respective data set group comprises samples having a time interval between adjacent samples, and wherein said waveform predictor has an order of not more than a maximum order substantially equal to the debabelizing interval of the computer system divided by said time interval.
16. A decoding method comprising the steps of:
receiving an encoded group representing information and obtaining therefrom data set information for respective square data sets of said information having scopes commensurate with or less than a corresponding critical band of computer capability, wherein said data set information corresponds to either prediction errors or a data set group,
generating a respective data set group for each data set represented by data set information corresponding to prediction errors by applying a predictor to the data set information, said predictor having an order greater than or equal to a minimum order, said minimum order equal to three, and
generating a replica of said information in response to said respective data set group for each of said data sets.
17. A decoding method according to claim 16 wherein, for a respective data set, said predictor is implemented by a digital filter having filter coefficients adapted in response to said data set group.
18. A decoding method according to claim 16 wherein said respective data set group comprises samples having a time interval between adjacent samples, and wherein said predictor has an order of not more than a maximum order substantially equal to the debabelizing interval of the computer system divided by said time interval.
19. A decoding method according to claim 15 or 18 wherein said replica of said information is generated by applying an inverse discrete transform to data set groups in said plurality of data sets, and wherein said minimum order is equal to 4, 6 and 8 for inverse discrete transform lengths of 512, 256 and 128, respectively.
20. A decoding method according to claim 19 wherein said inverse discrete transform substantially corresponds to either an evenly-stacked Time Domain Cancellation inverse transform or an oddly-stacked Time Domain Cancellation inverse transform.