Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1999-07-02
2001-11-20
Dorvil, Richemond (Department: 2641)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S244000, C704S251000, C704S231000
Reexamination Certificate
active
06321196
ABSTRACT:
TECHNICAL FIELD
The invention generally relates to speech recognition devices and methods and in particular to the phonetic spelling of words. Even more particularly it relates to using words from a vocabulary larger than the number of letters in the alphabet to indicate the letters of a phonetically spelled word. The invention also relates to using words from a vocabulary larger than the number of letters in the alphabet to select items from a list where each item is designated with one or more letters.
BACKGROUND OF THE INVENTION
Speech recognition devices have been developed with varying degrees of success. There is great variability in how different speakers pronounce words as well as variability in how an individual speaker pronounces words from one time to another. Current speech recognition technology has not yet been developed to the point of accommodating such variabilities to the extent with which a normal human listener can. For example, speech which is dictated into a mini-cassette recorder and then transcribed by a typist will typically have far fewer errors than if the same text is dictated directly to a current technology speech recognition computer program.
Several methods have been developed to assist the current technology devices in accommodating these variabilities, primarily through use of training. For example, the user may be asked to speak each new word at least once prior to using it. Or the speaker may be asked to read a list of frequently used words to the device. The speaker may be asked to monitor the recognized text and correct errors. All of these methods allow the recognition device to “learn” by adapting to the speaker's variability and in some cases variability between speakers. Nevertheless, it frequently occurs that the best approach for an unrecognized, difficult, or new word is for the speaker to spell it.
In other applications a speaker may select items from a list by saying the name of a letter associated with each item.
However, many of the letters sound very similar and may be confused for each other, even for human listeners. It is therefore known in the art to spell a word phonetically which is meant to indicate using a commonly understood word for each letter in the word to be spelled. For example, one may phonetically spell the work “key” by saying “kilo echo yankee.” One may also use a phonetic alphabet when selecting items from a list by using a word from the phonetic alphabet rather than saying the corresponding letter name.
A list of such words, one for each letter, arranged in alphabetical order is commonly known as a phonetic alphabet. Table 1 below lists an example of a phonetic alphabet.
TABLE 1
A
Alpa
N
November
B
Bravo
O
Oscar
C
Charlie
P
Papa
D
Delta
Q
Quebec
E
Echo
R
Romeo
F
Fox-trot
S
Sierra
G
Golf
T
Tango
H
Hotel
U
Uniform
I
India
V
Victor
J
Juliet
W
Whiskey
K
Kilo
X
Xray
L
Lima
Y
Yankee
M
Mike
Z
Zulu
Various such alphabets have been developed and used over the years with human listeners, primarily by the military, for clearly communicating over sometimes noisy or unreliable radio or telephone links. The developers of speech recognition devices and programs have likewise incorporated a phonetic spelling feature in their products for word spelling.
Other uses for phonetic spelling in association with speech recognition devices include communicating voice commands to a voice activated device as described by Basore et al. in U.S. Pat. No. 5,752,232, or to retrieve information from a directory, e.g. a telephone directory, in response to a phonetically spelled word as described by Dubnowski et al. in U.S. Pat. No. 4,164,025.
Phonetic spelling may be used to generate an audio output to a human listener in an audio response unit such as described by Barnett et al. in U.S. Pat. No. 4,653,100 and Silverman in U.S. Pat. No. 5,890,117. There is no speech recognition involved in this use of phonetic spelling which is the reverse process of speech generation.
In order to use a phonetic spelling feature, the speaker must have knowledge of the phonetic alphabet. This knowledge is easily learned in a military environment where, for example each signal core soldier is taught the phonetic alphabet as part of his signal core training. Ordinary users of speech recognition software have not been so trained and therefore keep a printed or handwritten list of the phonetic alphabet near their devices for use as necessary. Even so, it is awkward and slow for the ordinary user to visually search through the list for each phonetic word needed to phonetically spell a new word. As indicated above, the need to spell a word occurs more frequently when using current technology speech recognition devices than when dictating to a human transcriber because of the lesser accommodation to variations in pronunciation of the devices, further compounding the awkwardness.
OBJECTS AND SUMMARY OF THE INVENTION
It is therefore a principal object of the present invention to overcome the deficiencies of current technology speech recognition devices noted above by providing a phonetic spelling capability which is easier to use than those known in the art.
It is another object to provide a speech recognition apparatus capable of accepting phonetically spelled words by allowing the user to select phonetic words from a large vocabulary.
It is a further object to provide a method of recognizing phonetically spelled words using a large vocabulary of phonetic words.
It is another object to provide a method of selecting items from a list by speaking words from a large vocabulary of phonetic words.
It is yet another object to provide a computer program product for instructing a processor to recognize phonetically spelled words using a large vocabulary of phonetic words.
These and other objects are attained in accordance with one embodiment of the invention wherein there is provided a speech recognition apparatus, comprising, means for determining when a speaker gives an indication of a desire to phonetically spell a first word, means for recognizing a sequence of words selected form a vocabulary of greater than 26 words and spoken by the speaker after the indication, means for selecting a letter associated with each of the spoken words, and means for arranging the letters to form the first word.
In accordance with another embodiment of the invention there is provided a speech recognition apparatus, comprising, a display for showing items on a list, each item having a designated letter, means for recognizing one or more words selected from a vocabulary of greater than 26 words and spoken by a speaker, means for selecting a letter associated with each of the one or more words, and means for selecting the items on the list for which the designated letter matches the letter associated with each of the one or more words.
In accordance with another embodiment of the invention there is provided a method of recognizing speech, comprising the steps of, determining when a speaker gives an indication of a desire to phonetically spell a first word, recognizing a sequence of words selected from a vocabulary of greater than 26 words and spoken by the speaker after the indication, selecting a letter associated with each of the spoken words, and arranging the letters to form the first word.
In accordance with another embodiment of the invention there is provided a method of recognizing speech, comprising the steps of, displaying items on a list, each item having a designated letter, recognizing one or more words selected from a vocabulary of greater than 26 words and spoken by a speaker, selecting a letter associated with each of said one or more words; and selecting the items on the list for which the designated letter matches the letter associated with each of the one or more words.
In accordance with another embodiment of the invention there is provided a computer program product for instructing a processor to recognize speech, comprising, a computer readable medium, first program instruction means for instructing a processor to determine when a speaker gives and indication of a
Dorvil Richemond
International Business Machines - Corporation
Nolan Daniel A
Pivnichny John R.
LandOfFree
Phonetic spelling for speech recognition does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Phonetic spelling for speech recognition, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Phonetic spelling for speech recognition will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2571951