JPS58187054A - Collating system of voice and data mail talker - Google Patents

Abstract

PURPOSE:To preform discrimination talker, identification by subscriber's utterance of the same words as those registrated in advance from a subscriber and discriminating whether or not the voice quality of the said utterance is equal to the voice of talker registrated in advance. CONSTITUTION:A telephone set 6 is called by a bus 2 with a storage device 35 to inform that a confidential mail exists in a subscriber B, and the utterance of the registrated words in advance is requested. When the pass word and the set key are inputted to the subscriber B, the telephone set 6 is connected to a talker recognition device 34 in the route of ''1'' to discriminate whether or not the words uttered by the subscriber B is equal to the registrated words. The input/ output is done respectively for the key input through an I/O 31-1 to a cpu 32, for the call request from the storage device to the telephone set 6 through an I/O 31-3 to the cpu 32, th result of recognition from the talker recognition device through the I/O 31-3 to the cpu 32, and the switching of buses 1 and 2 through an I/O 31-2 from the cpu 32.

Description

[Detailed description of the invention] [Technical field to which the invention pertains] The present invention relates to a speaker verification method for confidential mail, etc.

[Prior art and its problems]

The conventional method will be explained with reference to FIG. In Figure 1, 1
is an exchange, 2 is a storage device, and 3 is a multi-frequency signal detector. When subscriber A wants to send a confidential message to B, A stores the message in the storage device 2 as a confidential mail via path ■. . FIG. 2 shows an example of the configuration of the storage device. 10-1～
10-n is line buffer, 11 is memory, 12 is cp
uX 13 is a disk controller, and 14 is a disk. For example, a message from someone is sent to line buffer 10-1.
, the memory 11, and the disk controller 13. Once stored in the storage device, the storage device calls the telephone 6, and if the telephone 6 is free, calls the telephone 6. If the telephone 6 answers, it is unknown whether the answerer is B or someone else, so a password is requested from the answerer. Subscriber B whose password from the responder has been registered in advance
If they match, the message is stored in the storage device and transmitted to the telephone 6. If they do not match, a request is made to the responder again, and if different passwords are received three times, for example, the connection to the telephone 6 is disconnected. To explain the connection form in Fig. 1, the first path (■) calls the telephone 6 from the storage device, and when the telephone 6 answers, a request is made to enter a password, and the multi-frequency signal detector (FB detection circuit) is connected to the pass (■). If the password matches that of subscriber B, the message from the person will be transmitted from the storage device to the telephone set 6 using the pass (3). To explain in more detail with reference to FIG. 3, dialing information, hooking information, etc. from the telephone are transmitted through line circuits 20-1, 20-n,
Ilo 21-1. PB detection circuit 20-n, Il
o Reach bip&-23 via 21-3. Control information necessary for the storage device reaches the CPU 24 via Ilo 21-4.21-5. Conversely, the control information from the storage device is from epLI・24 to Ilo 21-5.21-
4 to the CPU 43.

As mentioned above, in the conventional system, the subscriber has to key in or dial the password./(The password is often five digits long, and the subscriber has to memorize the password so that others do not know it.) It is necessary to memorize it secretly.Also, since it is input using common methods such as key-in and dialing, it is relatively easy for others to find out the password, as seen in the case of an American junior high school student who found out the password for a Canadian company and brought down the computer. easy to be

[Purpose of the invention]

The present invention has been made in view of these drawbacks. Therefore, instead of registering and detecting passwords using codes such as dials and keys manually, the subscriber utters the same words as the words registered in advance, This is a speaker verification method that performs speaker identification by determining whether the sound quality etc. is the same as that of a pre-registered speaker.

[Embodiments of the invention]

One embodiment of the present invention is shown in the block diagram of FIG.

First, the storage device 35 calls the telephone 6 using the pass (2), informs the subscriber B that the confidential mail is there, and requests the subscriber B to say a password, that is, a word registered in advance. For example, assume that B has registered his own name as "Toshiba Abe". B is the key with No1 password written on it,
Alternatively, when hooking is performed, the telephone 6 is connected to the speaker g fiber device 34 through the route ■, and the words uttered by B are 1yL.
Determine if the voice is equal to the one registered in advance.The key input or key power from the telephone 6 is sent to the CPU 3 through Ilo 31-1.
2, the call request from the storage device to the telephone 6 is sent to Ilo 31.
The line results from the speaker recognition device are also passed to the CPU 32 through l1031-3 to the CPU 32.
Switching is input and output from the epu through the Ilo 31-2.

An example of the configuration of a speaker recognition device is shown in FIG. amplifier 40,
High frequency pre-emphasis 41, automatic gain adjustment amplifier 42
, bandpass filter 43, full-wave rectifier 44, low-pass filter 45, storage device 46, intensity measurement circuit 47, pitch extraction circuit 48, short-time spectrum detection circuit 49, predicted thread count derivation circuit 50, formant frequency and band detection circuit 5
1. Nasopharyngeal articulation detection circuit 52, spectral correlation coefficient derivation circuit 53, timing and speech rate detection circuit 54, comparator 55
, dictionary (pre-registered audio signals) 56. Nine voices are passed through amplifiers and filters 40 to 45 and features are extracted using parameters 47 to 54 for speaker recognition.
By comparing it with a certain characteristic registered in the dictionary, 55 comparators are used to find out whether the speakers match or not.

[Other embodiments of the invention]

In the above method, −? In case the speaker's words are not sufficient for identification, recording multiple words will enable a more reliable speaker recognition method.

〔Effect of the invention〕

As described above, according to this method, it is possible to recognize the speaker without entering the password as a code, which not only simplifies the operation, but also makes it difficult to steal the password because it is not a code. This can also be prevented by having people speak multiple languages. In addition, it is possible to use not only a password but also dialing and other service keys 4 voice, and by combining these, it becomes possible to operate only by voice without dialing or key operations on the telephone, and a significant improvement on the operation side can be expected. The embodiment shown in FIG. 4 is not limited to this example; for example, the speaker recognition device and storage device may be the same, and the speaker recognition device shown in FIG. 5 is not limited to this example.

[Brief explanation of drawings]

Fig. 1 is a block diagram of an embodiment of a conventional audio/data storage device, Fig. 2 is a block diagram of the storage device, and Fig. 3 is a block diagram of an embodiment of a conventional audio/data storage device.
FIG. 4 is a block diagram of an embodiment of the present invention, and FIG. 5 is a block diagram of an embodiment of the speaker recognition device.・C9
11% 33...Memory, 34...Speaker recognition device, 35...Storage device. Agent Patent Attorney Kensuke Chika (and 1 other person) No. 1
Figure 2

Claims (2)

[Claims] (1) In a voice data storage system equipped with a means for accumulating voice or data and a means for recognizing a speaker, it is necessary to verify the speaker of confidential mail, registered mail, etc.
A voice/data mail speaker verification method characterized in that a speaker recognition device determines whether or not the speaker is coughing by uttering the voice. (2) The voice/data mail speaker verification method according to claim 1, characterized in that there is a plurality of voices.