JPS5967600A - Voice inputting system - 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 The present invention relates to a voice input method using an IC memory card.

There are two types of voice input methods: a speaker-specific voice input method in which speakers are limited, and a speaker-independent voice input method in which speakers are not selected.

Speaker-independent speech input methods generally have a small number of recognized words and a low recognition rate. On the other hand, a block diagram of the configuration of the specific speaker voice input method is shown in FIG. In this figure, 1 is the input voice of a specific speaker, 2 is a microphone, 3 is an analysis/feature extraction section, 4 is a switching section, 5 is a storage section, 6 is a recognition section, and T is a recognition output.

Input speech 1 of a specific speaker is sent to an analysis/feature extraction section 3 through a microphone 2, and extracted as speech feature parameters of the specific speaker. The voice characteristic parameters are first sent to the switching section 4 and then sent to the storage section 5 along the path shown by the dotted line in FIG.
is registered as a standard pattern for a specific speaker. In the case of voice input, the switching unit 4 is switched to the path shown by the solid line in FIG.
After being matched with the standard pattern, it is output as a recognition unit cuff, and is input to a host device not shown here.

In this way, in the specific speaker voice input method, only the standard pattern of the specific speaker is registered, so the recognition iiA! ! Although it has a relatively large number of words, it has the disadvantage that the number of speakers is limited.

The present invention eliminates these drawbacks and makes it possible to input voices of unspecified speakers while using the specific speaker voice input method. The present invention will be described in detail below based on the drawings.

Figure 2 is a block diagram showing an embodiment of the present invention. In this diagram, 1 is the input audio, 2 is the microphone, and 3
is an analysis/feature extraction unit, 5 is an accumulation unit, 6 is a recognition unit,
Up to this point, it is the same as in Figure 1. 8 is leader, 9 is I
C memory card is shown.

Next, the operation will be explained.

The input voice 1 is sent to the analysis/feature extraction unit 3 through the microphone 2, and the characteristic parameters of the speaker's voice are extracted.
It is sent to the recognition section 6. On the other hand, before voice input, speaker-specific voice characteristic parameters are written and stored in an EPROMIC, for example, by inserting into the reader 8 a portable IC memory card 9 that has been written and stored in the IC memory using a writing device (not shown). The information in the card 9 is read and stored in the storage section 5, and is used as a standard parameter unique to the speaker. By matching these standard parameters with the characteristic parameters of the cardholder's voice sent to the recognition unit 6, the input voice 1
is recognized and input as a recognition output T to a host device not shown here.

In addition, in this invention, a flexible disk is used.

It is possible to use magnetic disks, but they are expensive and
There are limitations as a portable device. Furthermore, this invention can be applied not only to voice input devices for specific speakers, but also to voice input devices for unspecified speakers, and it is also possible to ensure higher confidentiality by accumulating voiceprint information. be.

As explained above, by using an IC memory card that stores voice feature parameters for personal voice recognition, the present invention enables a voice input device for a specific speaker to be input to an unspecified number of speakers who own the IC memory card. Therefore, it is possible to input a large number of words with a high recognition rate. In addition, since the IC memory card is a small, lightweight, and economical portable card, voice input is possible in any location where the same or similar speaker-specific voice input device is installed. This has the advantage that a large number of people can economically use one speaker-specific voice input device.

[Brief explanation of drawings]

FIG. 1 is a diagram showing a conventional specific speaker voice input method, and FIG. 2 is a block diagram showing an embodiment of the present invention. In the figure, 1 is the input voice, 2 is the microphone, and 3 is the analysis/
Feature extraction section, 5 is storage section, 6 is recognition section, 7 is recognition output,
8 is the reader and 9 is the IC memory card.

Claims (1)

[Claims] The analysis/feature extraction unit of the specific speaker voice input device extracts a characteristic pattern from the input voice, and the recognition unit compares this characteristic pattern with the characteristic pattern of the individual who uttered the input voice, which has been stored in the storage unit in advance. In a voice input method that outputs voice characteristics such as phonemes, syllables, and words of the individual as a standard pattern for voice recognition,
By using a personal portable IC memory card stacked in the C memory element and loading the standard pattern into the storage section, voices of an unspecified number of speakers who own the IC memory card can be input to the specific speaker voice input device. A voice input method characterized by enabling input.