Harry Printz from South Orange, NJ

Us Patents

Augmentation Of Alternate Word Lists By Acoustic Confusability Criterion
view source
US Patent:

6754625, Jun 22, 2004
Filed:

Dec 26, 2000
Appl. No.:

09/746892
Inventors:

Peder Andreas Olsen - New York NY
Michael Alan Picheny - White Plains NY
Harry W. Printz - New York NY
Karthik Visweswariah - Jefferson Valley NY
Assignee:

International Business Machines Corporation - Armonk NY
International Classification:

G10L 1526
US Classification:

704235, 704251, 704255, 704257, 704270
Abstract:

There is provided a method for augmenting an alternate word list generated by a speech recognition system. The alternate word list includes at least one potentially correct word for replacing a wrongly decoded word. The method includes the step of identifying at least one acoustically confusable word with respect to the wrongly decoded word. The alternate word list is augmented with the at least one acoustically confusable word.

Determining And Using Acoustic Confusability, Acoustic Perplexity And Synthetic Acoustic Word Error Rate
view source
US Patent:

7219056, May 15, 2007
Filed:

Apr 19, 2001
Appl. No.:

09/838449
Inventors:

Scott Elliot Axelrod - Mt. Kisco NY, US
Peder Andreas Olsen - New York NY, US
Harry William Printz - San Francisco CA, US
Peter Vincent de Souza - San Jose CA, US
Assignee:

International Business Machines Corporation - Armonk NY
International Classification:

G10L 15/00
G06F 7/60
US Classification:

704235, 703 2, 704236, 704246, 704251
Abstract:

Two statistics are disclosed for determining the quality of language models. These statistics are called acoustic perplexity and the synthetic acoustic word error rate (SAWER), and they depend upon methods for computing the acoustic confusability of words. It is possible to substitute models of acoustic data in place of real acoustic data in order to determine acoustic confusability. An evaluation model is created, a synthesizer model is created, and a matrix is determined from the evaluation and synthesizer models. Each of the evaluation and synthesizer models is a hidden Markov model. Once the matrix is determined, a confusability calculation may be performed. Different methods are used to determine synthetic likelihoods. The confusability may be normalized and smoothed and methods are disclosed that increase the speed of performing the matrix inversion and the confusability calculation. A method for caching and reusing computations for similar words is disclosed.

Zero-Search, Zero-Memory Vector Quantization
view source
US Patent:

7729910, Jun 1, 2010
Filed:

Jun 25, 2004
Appl. No.:

10/877520
Inventors:

Harry Printz - New York NY, US
Assignee:

Agiletv Corporation - Menlo Park CA
International Classification:

G10L 15/00
US Classification:

704236
Abstract:

The invention comprises a method for lossy data compression, akin to vector quantization, in which there is no explicit codebook and no search, i. e. the codebook memory and associated search computation are eliminated. Some memory and computation are still required, but these are dramatically reduced, compared to systems that do not exploit this method. For this reason, both the memory and computation requirements of the method are exponentially smaller than comparable methods that do not exploit the invention. Because there is no explicit codebook to be stored or searched, no such codebook need be generated either. This makes the method well suited to adaptive coding schemes, where the compression system adapts to the statistics of the data presented for processing: both the complexity of the algorithm executed for adaptation, and the amount of data transmitted to synchronize the sender and receiver, are exponentially smaller than comparable existing methods.

Zero-Search, Zero-Memory Vector Quantization
view source
US Patent:

8185390, May 22, 2012
Filed:

Apr 23, 2009
Appl. No.:

12/429122
Inventors:

Harry Printz - New York NY, US
Assignee:

Promptu Systems Corporation - Menlo Park CA
International Classification:

G10L 15/00
US Classification:

704236
Abstract:

The invention comprises a method for lossy data compression, akin to vector quantization, in which there is no explicit codebook and no search, i. e. the codebook memory and associated search computation are eliminated. Some memory and computation are still required, but these are dramatically reduced, compared to systems that do not exploit this method. For this reason, both the memory and computation requirements of the method are exponentially smaller than comparable methods that do not exploit the invention. Because there is no explicit codebook to be stored or searched, no such codebook need be generated either. This makes the method well suited to adaptive coding schemes, where the compression system adapts to the statistics of the data presented for processing: both the complexity of the algorithm executed for adaptation, and the amount of data transmitted to synchronize the sender and receiver, are exponentially smaller than comparable existing methods.

Interval Width Update Process In The Arithmetic Coding Method
view source
US Patent:

55921620, Jan 7, 1997
Filed:

Mar 23, 1994
Appl. No.:

8/216741
Inventors:

Harry W. Printz - New York NY
Peter R. Stubley - Outremont, CA
Assignee:

Digital Equipment International, Ltd. - Fribourg
International Classification:

H03M 700
US Classification:

341107
Abstract:

The present invention relates to an interval width update process in arithmetic coding, characterized in that a set of values ={A[0],A[1],. . . ,A[r-1]}, is selected and the interval width is maintained as an index Wi in said set, a single table lookup simultaneously updates the interval width and supplies the augend and shift by performing the following operation: (Wi+1, Xi, Ri)=. function. "(Si, Wi) in which the function. function. " is implemented by a single table lookup, in which p(Si) and P(Si) are determined from Si, A[Wi] is determined from Wi, p(Si). multidot. A[Wi] and Ri=P(Si). multidot. A[Wi] are computed, the shift Xi necessary for representing p(Si). multidot. A[Wi]. multidot. 2. sup. Xi in is determined. Wi+1 is determined in such a way that A[Wi+1] is the best representative of p(Si). multidot. A[Wi]. multidot. 2. sup.

Method For Estimation Of Feature Gain And Training Starting Point For Maximum Entropy/Minimum Divergence Probability Models
view source
US Patent:

6049767, Apr 11, 2000
Filed:

Apr 30, 1998
Appl. No.:

9/070692
Inventors:

Harry W. Printz - New York NY
Assignee:

International Business Machines Corporation - Armonk NY
International Classification:

G10L 1100
US Classification:

704240
Abstract:

A method and apparatus for efficiently determining the gain of a feature function in a maximum entropy/minimum divergence probability model in a single pass through a training corpus. A method for determining the gain of a feature in such a model includes the steps of a selecting a set of evaluation points and determining the value of a function referred to as the gainsum derivative at each of the evaluation points. An approximation function which can be evaluated at substantially any point in a continuous domain is then selected based upon the discrete values of the gainsum derivative at the evaluation points. The approximation function is then employed to determine the argument value that maximizes an approximated gain function. The approximate gain value is then determined by evaluating the approximated gain function at this argument value. The apparatus of the present invention includes means for performing the steps of the disclosed method.

Code Point Update Device In The Arithmetic Coding Method
view source
US Patent:

55946747, Jan 14, 1997
Filed:

Oct 27, 1995
Appl. No.:

8/549467
Inventors:

Harry W. Printz - New York NY
Peter R. Stubley - Outremont, CA
Assignee:

Digital Equipment Corporation - Maynard MA
International Classification:

H03M 730
US Classification:

36471502
Abstract:

In a lossless compression algorithm, the left most point of a current interval, i. e. , the code point is updated by using a code point updating device which handles inputs identified by R and X, and uses two registers. The first of the two register is an interval width register, and the second is a code point register. The updating device comprises a grid of n. times. n cells chosen from four basic component cells, i. e. , (i) above diagonal cells designated "ad" and located above the principal diagonal of the grid, (ii) diagonal cells designated "d" and located on the principal diagonal, incorporating and adder (31) and a multiplexer (30) and at least one register (32); (iii) a single cell designated "fd" for "final diagonal", located at the bottom end of the diagonal and incorporating an adder (31) a multiplexer (30) and at least one register (32), and (iv) output retiming cells designated "or", incorporating a register (35).