Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
2002-08-14
2003-03-18
McFadden, Susan (Department: 2654)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
C704S270000, C434S156000, C434S167000
Reexamination Certificate
active
06535853
ABSTRACT:
BACKGROUND OF INVENTION
The present invention relates to systems and advanced methods for the detection of various forms of dyslexia, and more particularly to such systems and methods incorporated in existing handwriting and speech recognition systems.
DESCRIPTION RELATIVE TO THE PRIOR ART
Dyslexia is a general term for a family of learning disabilities. Its symptoms include problems in expressive or receptive oral or written language. Derived from the Greek words “dys” (poor or inadequate) and “lexis” (words or language). Dyslexia affects reading, spelling, writing, memory and concentration, and sometimes math, music, foreign languages and self-organization. Dyslexia and other related learning disabilities affect about 15% of the population.
There is a consensus as to the need for early detection and treatment. The symptoms can be difficult to detect, however, especially at an early age. In pre-schoolers these symptoms may include the following:
Later than expected learning to speak clearly.
Jumbling phrases, e.g. ‘cobbler's club’ for ‘toddler's club’, ‘teddy-dare’ for ‘teddy-bear’.
Use of substitute words or ‘near misses’.
Mislabeling e.g. lampshade for lamp post.
Speaking with lisp—‘duckth’ for ‘ducks’.
Inability to remember the label for known objects, e.g. colours.
Confusing directional words, e.g. ‘up/down’ or ‘in/out’.
Excessive tripping, bumping and falling over nothing.
Enhanced creativity—often good at drawing—good sense of colour.
Obvious ‘good’ and ‘bad’ days for no apparent reason.
Having aptitude for constructional or technical toys, e.g. bricks, puzzles, lego blocks, control box for TV and video, computer keyboards.
Enjoying being read to but showing no interest in letters or words.
Difficulty learning nursery rhymes.
Finding difficulty with rhyming words, e.g. ‘cat mat fat’.
Finding difficulty with odd-one-out, e.g. ‘cat mat pig fat’.
New research has identified three subtypes of dyslexia identified as dyseidesia, dysphonesia, and dysphoneidesia. Dyseidesia is a deficit in the ability to perceive whole words with their sounds. This subtype mainly affects the reading of non-phonetically regular words, such as laugh. Dysphonesia is a deficit in word analysis skills. People with this subtype have trouble with unfamiliar words as well as word association, as in the difference between home and house. Dysphonedeisia is a combination of the previous two subtypes.
The current invention is based on the well-known tendencies of dyslexics to perceive characters in unique ways, both in the perception of the characters, and in attempts to reproduce those characters by drawing the characters as they see them. It is further based on the inability of dyslexics to utilize normal discretionary skills to recognize these individual character groupings when they are assembled into words and to pronounce them properly.
The English language offers special problems in pronunciation due to its varied roots in distinct language groups. English is not a phonetic language, and the same combinations of letters are pronounced differently in different words.
For example, the combination “ough” has a different pronunciation in the words “through”, “though”, “thought”, “tough”, and “plough”. Similarly, the word “bow” is pronounced differently when used in the sense of bending at the waist, as compared to meaning a knot in a ribbon.
This invention is implemented by means of various software and hardware products that are well developed and in wide use at the present time. These include handwritten character recognition systems, both printed and cursive, as well as text to speech systems and automatic speech recognition systems. These systems all have one characteristic in common. The systems compare inputs, both handwritten and spoken, to a library of acceptable, ideal or target models, and reject those characters that do not conform to any recognized models in their various libraries. These rejected or failed input attempts, both written and spoken, are stored on magnetic system disks for later analysis.
The present invention expands this technique to include not only recognizable legitimate inputs, but also illegitimate or rejected inputs. When the test subject of one of these systems generates a significant number of these failed inputs, the invention records these in memory associated with the system″s handwriting recognition and speech recognition systems and creates a statistical profile of the errors, thus enabling the data to produce a report warning that the test subject may be displaying characteristics associated with a specific form of dyslexia.
The stored inputs have the added advantage of being able to be printed in hard copy form, displayed on a computer screen and or simultaneously listened to on the systems associated audio system.
This invention has the great advantage that it does not require any advances in the basic technologies of handwritten character recognition, text to speech, or automatic speech recognition systems.
The current invention can be incorporated into games currently available on computer and INTERNET systems for young children of pre-school and early school age, thus providing the early warning so urgently needed.
SUMMARY OF INVENTION
It is an object of the present invention to provide a method to detect the inability of a test subject to reproduce letters of the alphabet, and to further detect the inability of the test subject to properly pronounce the test words.
In accordance with one aspect of the method, a library of test words is constructed, wherein for each word, the library includes a sound recording of the correct pronunciation of the word, a written representation of the word, and the identity of each phoneme contained therein; In accordance with another aspect of the method, the system displays the written representation of one of the words contained within the library, and the subject is requested to pronounce the word displayed.
In accordance with a third aspect of the method, the pronounced word is pronounced by the subject, and recorded as a response sound record.
In accordance with a fourth aspect of the method, the response sound record is compared with the sound record of the test word, and the response is recorded as properly or improperly pronounced, together with the identity of the phonemes contained in the test word.
In accordance with a fifth aspect of the invention, the process is repeated until a statistically significant number of test words has been processed.
In accordance with a sixth aspect of the invention, a report is generated summarizing the number of proper and improper pronunciations, the words improperly pronounced, and the phonemes contained in each such word.
In accordance with a seventh aspect of the invention, the comparing of the response sound record of the sound record of the test word is done by currently used voice recognition techniques.
In accordance with an eighth aspect of the present method the test words displayed further are made up of words containing phonemes which are different, but have identical spellings.
In accordance with a ninth aspect of the method, the voice recognition analysis further includes breaking down the response record into phonemes, wherein each sound record in the library of test words further includes a collection of separate sound bytes, each of which represents a separate phoneme of the corresponding test word.
In accordance with a tenth aspect of the present invention, the report includes an audio record of each improperly pronounced word, which may be output to an output device, such as a loudspeaker, during a review of the report.
REFERENCES:
patent: 5596698 (1997-01-01), Morgan
patent: 5920838 (1999-07-01), Mostow et al.
patent: 6468084 (2002-10-01), MacMillan
McFadden Susan
White Mark P.
LandOfFree
System and method for dyslexia detection by analyzing spoken... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for dyslexia detection by analyzing spoken..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for dyslexia detection by analyzing spoken... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3085723