JPS6037501B2 - voice input typewriter - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an improvement in a voice input typewriter for inputting character information by voice.

The work of inputting character information into a computer is an important work in information processing work, and improving its efficiency is considered essential for improving the efficiency of the entire information processing work.

Particularly in our country, since the use of typewriters is not common, a large amount of training is required to train new character data entry operators. Therefore, it is strongly desired to develop an easy-to-use character data input means that can replace the conventional keyboard. Attempts were made to develop a voice input typewriter as a promising candidate for this purpose. A voice input typewriter is a device that converts human voice into character data codes through a pattern recognition process in which the voice uttered by a person is analyzed and identified by an electronic circuit. However, recognizing natural speech is an extremely difficult problem, and it is said that there is almost no possibility that a practical speech typewriter will be realized in the near future. The biggest reason why it is difficult to recognize speech patterns is that it is difficult to identify consonants. For example, no effective discriminative features can be obtained between pa, ta, forced, etc., even by detailed analysis.
This can be said to be the most difficult example to identify. This is the consonant lpl, ltl
, lkl are difficult to distinguish from each other. The present invention avoids the problem of consonant recognition by inputting consonant information that is difficult to recognize using another means such as a keyboard, thereby realizing and providing a voice input typewriter that operates accurately and is an inexpensive means for inputting character information. It is an object.

The voice input typewriter according to the present invention mainly includes a manual input means such as a keyboard for inputting consonant information, and a function to recognize and process separately input voice patterns within the range of phonemes having consonants designated as the consonant information. It is composed of a voice recognition section having a.

According to such a configuration, complicated processing for consonant recognition is not required, so that the overall configuration can be made compact and inexpensive.

Furthermore, even if a keyboard is used to input consonant information, it is sufficient to be able to specify the consonant, so the number of keys may be small, and therefore operation is easy. The principle of the embodiment will be explained below based on the embodiment.

FIG. 1 is a diagram showing an embodiment of the present invention. For example, when the user of this voice typewriter wants to input the character "sa", he/she must specify the letter "sa" from the keyboard section 10' using the key marked with the letter "s", and at the same time input the character "sa" from the microphone 11. Input the voice “sa”. The keyboard section 10 is constructed as shown in FIG. 2, for example. That is, a key marked with the letter "a" specifies a vowel and a phoneme similar to the vowel. For example, A, I, U, E, O, Y, U.
, wa, and n are specified by this key. These are called vowel groups. Further, the keys marked with the letter "k" specify power running such as force, ki, ku, ke, ko, kya, cu, kyo. Similarly, g (ga sound), s (sa sound), z
(za sound) t (ta sound), d (da sound), n (na sound)
, h (C sound), p (P sound), b (B sound), m (M sound), r (R sound), etc. are provided. Apart from the punctuation marks “・”, “.”, and “, just like a normal typewriter,
Keys such as "Line feed" and "Blank" may also be provided, as well as keys for specifying numbers as necessary.When entering numerical data, press this "number" key to enter "3" or "8". Say something like this and input it into the microphone 11. When a key is pressed, the keyboard section 11 outputs the following code as a consonant signal c.
Output as . a → 0, k → 1, g → 2, s → 3, z
→4, t→5, d→6, n→7, h→8, p→9, b→
10, m → 11, r → 12, number → 13, ・ → 14,.

→15, Line break →10 Blank →17 Decoding circuit 2 decodes the consonant signal c given from the keyboard section 10, and when the consonant signal c has a value in the range of 0 to 13, the selection signal c
Only the corresponding one among oc and 3 is set to 1, and the others are set to 0.

For example, when entering the character “sa”, press “” on the keyboard.
Since the "s" key is pressed, the consonant signal becomes c=3. Therefore, only the selection signal c3 becomes 1 and the others become 0. The speech recognition section includes standard pattern storage sections 30 to 43, microphone 11,
Analysis section 50, input pattern buffer 60, matching section 7
0, and a determination unit 80.

The standard pattern storage units 30, 31,...43 have the following information:
Respectively, vowel group, power group, ..., numbers (0~
9), the standard voice patterns are stored.

For example, the block indicated by the reference numeral 33 stores standard patterns for the ``s'' sounds (sa, shi, su, se, so, sha, shu, sho). This standard pattern is, for example, “IEE Transaction VOL.ASSp-8, No.1 (
``Dynamic Programming Algorithm Optimization For Spoken Word Recognition'' (February 197) [rlEEE Transact
ion VoL. ASSp-26, No. 1 (1978
February) “Dy Misaki mic Pro carp ammlng
Algori ratio optimization method on for
word recognition
As described in the paper titled ``''], the time series B=b,Q,...,bi... of the feature vector bi.
...expressed as bJ {1}. The standard pattern storage units 30 to 43 each store all necessary standard patterns of speech in a form such as m-type. When the selection signal c3 is 1 and the others are 0, only the standard pattern of the sa-go sound stored in the block indicated by the reference numeral 33 is selected from the bus line b.
are output sequentially. On the other hand, the sound "sa" inputted from the microphone 11 is spectrum-analyzed by the analyzer 50, and time-sampled at a constant frame period to create a feature vector a representing the spectrum envelope at each point in time.
Time series A of i=a,,a2,...ai,...a・
It is input to the buffer 601 in the format {2}. This pattern A is called an input pattern. In the matching section 70, the input pattern A stored in the input pattern buffer 60 is connected to the standard pattern storage section 3 via the bus line b.
From 3 onwards, a comparison operation is performed with each of the standard pattern groups of the sa-gradation sound given in the form of Equation 11.

This comparison operation is performed, for example, by the DP matching method described in the above-mentioned paper, and as a result, a time-normalized distance is output from the signal line d. The determining section 80 compares the time-normalized distances between the input pattern A calculated by the matching section 70 and each of the standard patterns of the sa-gradation sound, and determines the minimum standard pattern. When the input voice is "sa", the time normalized distance from the standard pattern of "sa" is naturally the minimum. In this way, the input speech is recognized as sa. This recognition result is output as a signal v. If only vowel information is given as a recognition result, character information can be completely synthesized by combining it with the consonant signal c, so the recognition result signal v is determined as follows. In other words, regardless of the number of rows, A step → 0, A step → 1, U step → 2, E step → 3, O step → 4, Y step → 6, U step → 6, Yo step → 7, Wa step → Let the code 8-9 be the recognition result signal v.

In the case of numbers, codes 0 to 9 are output as they are as recognition result signals v for each of 0 to 9.
The main character code synthesis section 90 synthesizes the main character code using the consonant signal c given from the keyboard section 10 and the recognition result signal v given from the determination section 80. FIG. 3 shows an example in which the above-described main character code synthesis section is constructed using a dedicated memory.

2 becomes (c, v) by the consonant signal c and the recognition result signal v
For example, at address (7, 3), the character “N” (c=7) and vert (v=3) is specified.
Memorize the character code. Also, address c=13 (
13,0) to (13,9) are stored with real character codes of numbers "0" to "9". c=14,15,16,1
Address number 7 has “・”, “.”, and “line feed” regardless of v.
, and “blank” codes are programmed. In this way, the code of the character specified by the consonant signal c and the recognition result signal v is read out and output as an output signal r. This output signal can be connected to a typewriter printing mechanism, paper tape puncher, computer, etc. as desired. Although the principle of the present invention has been explained above based on examples, these descriptions do not limit the scope of the present invention.

As the consonant information input means, a keyboard or the like may be used. Furthermore, when using a keyboard, the key arrangement is not limited to that shown in FIG. Specifically, the voice recognition section may also be constituted by a microcontroller or the like, and the recognition principle may also be different from that of the above embodiment.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an embodiment of the present invention. 10 is a keyboard section, 11 is a microphone, 20 is a decoding circuit, 30 to 43 are standard pattern storage sections, 5 river analysis sections, 60 is an input pattern buffer, 70 is a matching section,
80 is a determination section, and 90 is a character code synthesis section. FIG. 2 is a diagram showing an example of the configuration of the keyboard section. FIG. 3 is a diagram showing an example of the configuration of the character code synthesis section.
91 is a read-only memory. Younger brother's younger brother 2 drawings 3 drawings

Claims (1)

[Claims] 1. A voice input typewriter, comprising at least a manual input means for inputting consonant information, and a voice recognition unit that recognizes and processes the input pattern within the range of phonemes having consonants specified by the consonant information. A voice input typewriter characterized by: