CONVERSION OF PRINTED TEXT INTO SYNTHETIC SPEECH
First Claim
Patent Images
1. A system for converting printed data into artificial speech which comprises, in combination:
- means for converting printed material into alphanumeric signal data;
means for analyzing said alpha-numeric data to determine a syntactical phrase category for individual word designations therein;
means responsive to said syntactical data and to stored rules for assigning parametric values, including pause, stress, duration, pitch, and intensity values, for individual words in said alpha-numeric data;
means responsive to said parametric valves for assigning individual phonetic values, including phoneme, duration, and pitch values, to said word intervals; and
means responsive to said phonetic values for developing a continuous description of said alpha-numeric data as the energizing source of control signals for a speech synthesizer.
0 Assignments
0 Petitions
Accused Products
Abstract
A system is disclosed for converting printed text into speech sounds. Text is converted to alpha-numeric signal data, for example, by a scanner and dictionary lookup. Syntax of the input information is then analyzed to determine the proper phrase category, e.g., subject, verb, object, etc., of word intervals, and to assign pause, stress, duration, pitch and intensity values to the words. From these data a phonetic description of each word is found in a stored dictionary, modified by the accumulated data, and used to prepare synthesizer control signals.
-
Citations
9 Claims
-
1. A system for converting printed data into artificial speech which comprises, in combination:
- means for converting printed material into alphanumeric signal data;
means for analyzing said alpha-numeric data to determine a syntactical phrase category for individual word designations therein;
means responsive to said syntactical data and to stored rules for assigning parametric values, including pause, stress, duration, pitch, and intensity values, for individual words in said alpha-numeric data;
means responsive to said parametric valves for assigning individual phonetic values, including phoneme, duration, and pitch values, to said word intervals; and
means responsive to said phonetic values for developing a continuous description of said alpha-numeric data as the energizing source of control signals for a speech synthesizer.
- means for converting printed material into alphanumeric signal data;
-
2. A system as defined in claim 1, in further combination with:
- means for storing said parametric values for selected printed data; and
means responsive to a command signal for delivering said stored parametric values to said means for assigning individual phonetic values thereto.
- means for storing said parametric values for selected printed data; and
-
3. A system for converting printed text into speech sounds, which comprises:
- means for converting printed text material into alpha-numeric signal data;
means for examining the syntax of said alpha-numeric data to assign an appropriate phrase category to each word in said text material;
means responsive to said phrase category assignments for assigning a pause probability value For each word boundary in said material;
means responsive to said alpha-numeric data, to said phrase category assignment values, and to said pause probability assignments for developing signals representative of pauses in said text material;
means responsive to said alpha-numeric data, to the syntax of said data, and to said pause signals, for developing signals representative of duration, pitch, and intensity values of each word in said material;
means responsive to said assigned values for selecting a corresponding phonetic description of each word of said material from a dictionary of stored phonetic descriptions of words; and
means responsive to stored rules for converting said selected phonetic descriptions into synthesizer control signals.
- means for converting printed text material into alpha-numeric signal data;
-
4. A system as defined in claim 3, wherein said means for developing signals representative of pauses in said text material comprises:
- means for representing said pause probability values for each word as a proportioned value signal;
means for detecting the largest of said probability value signals in a selected sequence of alpha-numeric signal data; and
means for developing a signal representative of said largest detected signal value as a representation of a pause of said text material.
- means for representing said pause probability values for each word as a proportioned value signal;
-
5. A system as defined in claim 3, wherein said means for developing signals representative of pauses in said text material comprises:
- means for representing said pause probability value for each word as a proportioned value signal;
means for establishing a variable threshold whose value is related to the length of selected alpha-numeric signal sequences; and
means for developing a pause representative signal for the word denoted by that proportional signal value that exceeds said threshold value.
- means for representing said pause probability value for each word as a proportioned value signal;
-
6. A system for converting English text into artificial speech control signals, which comprises, in combination:
- a dictionary of signal information corresponding to a selected large vocabulary of words and phrases and variations of them;
means responsive to signal counterparts of printed text and to said dictionary signal information for assigning a syntactical grammatical phrase category to each individual word of said text;
means responsive to said assigned categories, and to stored rules for developing signals representative of desired pauses at boundaries between words, phrases, and clauses in said printed text;
means responsive to said dictionary signal information, said pause signals, and to stored rules, for developing signals representative of timing control values, pitch contour values, and intensity values for said individual words; and
means responsive to said dictionary signal information and to said timing, pitch, and intensity value signals for developing a continuous signal description of said English text as the energizing source of control signals for a speech synthesizer.
- a dictionary of signal information corresponding to a selected large vocabulary of words and phrases and variations of them;
-
7. A system as defined in claim 6, in further combination with:
- means for storing said continuous signal description of said English text, and command means for selectively withdrawing said stored signal description for delivery to a speech synthesizer.
-
8. A system for converting printed data into artificial speech, which comprises:
- means for converting printed data into textual signals;
means for analyzing said textual signals to develop a syntactical phrase category signal for individual word designations therein;
means responsive to said syntactical phrase category signals and to stored rules for assigning pause, stress, duration, pitch, and intensity values to individual words in said material;
means for storing signals for a vocabulary of phonetic description of words and phrases, each in a number of different variations of pronunciation, usage, and grammatical form;
means responsive to said assigned word values for selecting a corresponding phonetic description for each word of said printed material from said stored vocabulary signals;
means responsive to said selected phonetic descriptions for developing a continuous signal description representative of said printed material as the energizing source of control signals for a speech synthesizer; and
a speech synthesizer energized by said continuous signal description.
- means for converting printed data into textual signals;
-
9. A method for converting printed material into artificial speech control signals which comprises the steps of:
- converting printed material into alpha-numeric signals, analyzing said alpha-numeric signals to assign a syntactical phrase category for individual word designations therein, assigning parametric values, including pause, stress, duration, pitch, and intensity values, for individual words in said alpha-numeric signals in dependence on said assigned syntactical phrase categories and on stored rules, assigning individual phonetic values to said word intervals in dependence on said assigned parametric values, and modifying said alpha-numeric signals according to said parametric values, said phonetic values, and said stored values, to produce a continuous signal descriptive of said printed material for use as energizing source of control signals for a speech synthesizer.
Specification