JPS60182495A - Japanese language voice input unit - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] <Technical Field of the Invention> The present invention is a system for a specific speaker in which a standard pattern for each individual is registered in advance, the input speech is recognized in monosyllable units, and the result is output. The present invention relates to an improvement of a Japanese speech input device, and more particularly to a Japanese speech input device having a consonant mode in which a consonant is automatically detected or a consonant is not detected.

<Technical background of the invention and its problems> In general, when recognizing Japanese speech in monosyllable units, a method for processing consonants is required.There are various ways to process consonants, but the following are typical. Things include the following:

■ Specify consonants, consonants, etc. from the keyboard.

In this method, for example, press the lowercase letter key and say ``tsu'' to input ``tsu.'' When entering ``school,'' when saying ``ga tsukotsu,'' the two underlined characters are For eyes, press the lowercase letter key to make the sound.

■ Pronounce the consonant as "tsu" and automatically convert it into "tsu" in later dictionary processing.

For example, this method was previously proposed by the present applicant and is shown in Japanese Patent Application Laid-Open No. 195900/1983,
If you operate the conversion key after saying this, it will automatically be converted and input as "Pakakko".

■ Equipped with a means to detect consonant sounds, eliminating restrictions on vocalization.

In this method, for example, when the user utters ``ga ko'', the underlined part is detected as a silent section, and this section is input as ``kakko'' as the consonant detection section.

Each of the above-mentioned consonant processing methods has the following problems.

That is, in the method (■), it is necessary to operate the keyboard each time a consonant is produced. In addition, in method (2), uppercase letters ``1'' and lowercase letters ``1'' are confused, so auxiliary input means and operations for specifying consonant sounds are not necessary, but there is a problem that the number of conversion results increases. For example, when a child pronounces Paatkankanpa, he or she cannot distinguish between ``Atsukan'' (hot sake) and ``Akan'' (highlight).

The method described in ``Kubida■'' is easy to use once the narrator gets used to the device, but if he or she is not used to it, the problem is that the speaker cannot speak in a certain way and consonants are inserted in unexpected places.

<Object of the Invention> The present invention has been made with the purpose of providing a Japanese voice input device that eliminates the above-mentioned conventional problems, and in order to achieve this purpose, the Japanese voice input device of the present invention is The device is configured to include a consonant detection means for detecting a consonant, and a consonant mode storage unit that stores a consonant processing mode for determining whether or not the consonant detection means automatically detects the consonant. With such a configuration, the Japanese speech input device of the present invention allows the device to select either the conventional consonant processing method (■) or the consonant processing method (■) as the consonant detection mode at the time of registration. For people who are not accustomed to using the device, we will use the consonant processing method (■) to prevent consonant sounds from entering unnecessary places, and for people who have become accustomed to the device and can speak in a certain way,
Optimal consonant processing can be achieved by setting the consonant processing mode (3) in which consonant automatic detection is performed. According to the embodiment of the present invention, the selected consonant processing mode is stored in the same external storage means as the standard pattern, so that the optimal consonant processing mode can be selected from the next time simply by reading the standard pattern from the external storage means. The settings are configured to be entered.

<Embodiment of the Invention> Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.

In FIG. 1, 1 is a microphone for inputting voice, 2 is an amplifier for amplifying the voice input, and 3 is a consonant detecting section for detecting a consonant in automatic consonant detection mode. For example, it is configured to detect a silent section of a voice input, and if the length of the detected silent section is less than a certain threshold value, it is detected as a consonant. In order to detect the consonant more reliably, the consonant detecting unit 3 detects the continuation time of a vowel included in the input voice, and determines the speaking rate of the input voice based on this vowel continuation time. is estimated, and a threshold for the length of the silent section of the input voice is calculated based on this estimated speech rate. If the length of the silent section of the detected input voice is less than the calculated threshold, a consonant is detected. It may also be configured to detect as follows.

Reference numeral 4 denotes a consonant mode memory which stores the consonant detection mode as a flag bit, and when the consonant mode memory 4 stores, for example, bit information R, +1, the consonant detecting section 3 is enabled. When the bit information ++011 is stored, the consonant detection section 3 is set to a non-operating state (
The configuration is such that a consonant processing mode is established in which consonant detection is not performed.

5 is a syllable recognition unit that recognizes syllables based on the input voice;
6 is a standard pattern memory for storing standard patterns registered in advance; 7 is a CPU for controlling each part of the device; and 8 is an external memory for storing registered standard patterns, consonant processing modes, etc., such as a floppy disk. The memory information 9 is transferred between the consonant mode memory 4 and the standard pattern memory 6 by controlling the CPU 7.

9 uses the syllable candidates recognized by the recognition unit 5 to
This dictionary matching unit 9 performs dictionary matching of the syllable string created in step 7 and outputs the most likely candidate. Note that Vcl is configured so that this replacement function functions only when the bit information '0' is stored in the consonant mode memory 4.

10 is a dictionary memory for storing the contents of the dictionary; 11 is an input section having key manual means for instructing the consonant processing mode, selecting the next candidate for the recognition result, etc.; and 12 is a display section for displaying the recognition results, etc. .

Next, the operation of the device configured as described above is illustrated in FIGS. 2 to 5.
The explanation will be given according to the operation flowchart shown in the figure.

Figure 2 shows the operational flow when an operator (speaker) uses the device shown in Figure 1 above for the first time. After registering the pattern etc. (step n+
), voice input operation will be performed (step n2)
.

FIG. 3 is a processing flow showing the registration operation in step n1 in detail.

When registering an operator-specific standard pattern, the operator first selects whether the consonant processing mode is a mode that automatically detects consonants or a mode that does not detect consonants, and then inputs the selected consonant processing mode information. Input section 1
Input from step 1 (step n++). When a mode that automatically detects a consonant is selected by such an operation, bit information ++, +1 is stored in the consonant mode memory 4, and when a mode that does not detect a consonant is selected, the bit information ++0)) is a consonant. It is stored in the mode memory 4.

. Next, when the operator vocally inputs a predetermined sound, for example, a word, the recognition unit 5 creates a standard pattern for each unit clause and stores it in the standard pattern memory 6 (n13
). Thereafter, according to an instruction from the input unit 11, the standard pattern created and stored in the standard pattern memory 6 is transferred and stored in an external memory 8 such as a floppy disk device together with the consonant processing mode information stored in the consonant mode memory 4. (n13).

As described above, the registration operation of step n1 shown in FIG. 2 is executed.

Next, the input operation at step n2 shown in FIG. 2 will be explained.

FIG. 4 is a processing flow showing the input operation in step n2 in detail.

Now, the voice input uttered by the operator is converted into a voice signal by the microphone 1, and then amplified by the amplifier 2 and given to the consonant detector 3. The operation of this consonant detection unit 3 is selected according to the consonant processing mode information stored in the consonant mode memory 4 (step n21), and for example, in the consonant automatic detection mode in which the consonant processing mode information is set to °゛l''. If there is a consonant, the consonant detection unit 3 operates to detect a consonant (n22), and if a consonant is detected in the input audio signal (n23), a lowercase letter is inserted at that position (n24). After monosyllable recognition is performed in the recognition unit 5, C
PU7. Conversion processing is executed using the dictionary collation unit 9 and dictionary memory IO (n25).

If no consonant is detected in step n23, conversion processing is performed based on the middle tone recognition result of the recognition unit 5.

Further, in step n2+VC, if the consonant processing mode information stored in the consonant mode memory 4 is a mode in which consonants are not detected, which is set to 0'', the consonant detecting section 3 does not operate and the input voice remains unchanged. The recognition unit 5 recognizes each single syllable and moves on to the subsequent conversion operation (slurry n2+, n2
5).

In this conversion operation, the dictionary matching unit 9 additionally performs dictionary matching by replacing the uppercase letter ``p'' in the syllable candidate string with the lowercase letter ゛゛゛. Performs a dictionary lookup.

As described above, the consonant processing in the consonant automatic detection mode and the consonant non-detection mode are respectively executed according to the storage state of the consonant mode memory 4.

When the above-mentioned apparatus is used for the second time or later, the operation flow e shown in FIG. 5 is followed.

That is, first, the standard pattern stored in the external storage memory 8 is read out and stored in the standard pattern memory 6 in the main body of the apparatus (step n3). At this time, the consonant processing mode information stored in the external storage memory 8 mentioned above is also read out and stored in the consonant mode memory 4 within the main body of the apparatus.

The input operation of the next step n4 is performed, but this operation is the same as the operation shown in FIG. A mode in which no consonant is detected is automatically selected, and voice input can be performed in a consonant processing mode suitable for the operator itself.

As described above, when registering a speech standard pattern, the mode of consonant processing is selected as either no consonant detection or automatic consonant detection, and that mode is stored in the same external storage device as the standard pattern. By doing this, the next time the operator inputs, it becomes possible to input in a consonant processing mode suitable for the operator.

Note that in the above embodiment, the input unit 11 such as a keyboard
Enter registration mode even when inputting according to instructions from
It is more desirable to be able to forcibly escape from the registration mode even during registration.

<Effects of the Invention> As described above, according to the present invention, there is provided a Japanese speech input device in which standard patterns for each individual are registered in advance, input speech is recognized in units of monosyllables, and the results are output. The method is configured to include a consonant detection means for detecting a consonant, and a consonant mode storage section that stores a consonant processing mode indicating whether to perform consonant automatic detection by the consonant detecting means or not to perform consonant detection. For operators who are not familiar with the device, the mode is set to detect consonants in order to prevent consonants from being inserted where they are not needed, and for operators who are accustomed to the device and can speak in a certain way, the consonant mode is used. It is possible to set the mode to automatically detect the consonants, and perform the optimal consonant processing that suits the operator using the device.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, FIG. 2 is a diagram showing the operational flow at the time of first use, FIG. 3 is a diagram showing the operational flow at the time of registration, and FIG. FIG. 5 is a diagram showing the operation flow for the second and subsequent times. 3... Consonant detection unit, 4... Consonant mode memory, 5...
Recognition unit, 6... Standard pattern memory, 7... CPU
, 8...external memory, 9...dictionary collation unit, 10...
・Dictionary memory, 1'...input section. ：、( ``Representative Patent Attorney Aihiko Fukushi (and 2 others) 9

Claims (1)

[Scope of Claims] l A Japanese speech input device in which a standard pattern for each individual is registered in advance, input speech is recognized in monosyllable units, and a result is output, and a consonant detection means for detecting a consonant. A Japanese speech input device comprising: a consonant mode storage section that stores a consonant processing mode for determining whether to perform consonant automatic detection by the consonant detecting means or not to perform consonant detection. 2. The consonant processing mode is selected at the time of registration, and the selected consonant processing mode is stored in the same external storage means as the standard pattern, and at the next input, the consonant processing mode stored in the external storage means is changed to the consonant processing mode. 2. The Japanese voice input device according to claim 1, wherein the Japanese voice input device is configured to be written in mode storage means.