JPH05316198A - Caller recognition telephone set - Google Patents

Abstract

PURPOSE:To eliminate the need for limiting an object audio to a content of a message learned in advance in the case of talker recognition. CONSTITUTION:When a talker recognition section 13 of a caller recognition telephone set 10 uses a neural network to recognize a talker, the section 13 receives a series of vectors representing an outline of a spectrum for a short period of time, a series of network outputs is totallized by sum, product, majority decision or the like for the result of recognition based on each output to obtain one recognition result, and number of learning data for the network is decreased by clustering.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a caller identification telephone device which can make a call only to a speaker registered in advance in order to prevent a mischievous telephone or a wrong telephone from being bothered.

[0002]

2. Description of the Related Art Conventionally, Japanese Patent Laid-Open No. 59-208965, "Calling System for Telephone", Japanese Patent Laid-Open No. 60-239163, "Automatic Answering Telephone System", Japanese Patent Laid-Open No. 2-55448, "Voice Dialing System", Japanese Patent Laid-Open No. 3-35648.
A "nuisance call prevention device" and the like have been proposed. In these conventional techniques, there is little description about a specific recognition method,
Even if it is described, the description is only to the extent shown below.

A feature amount relating to a speaker included in the input voice is extracted, a distance between the extracted data and dictionary data extracted in advance by a similar method is calculated (similarity calculation), and this calculation is performed. Based on the result, it is determined whether the input voice is the voice of the registered speaker, and if the input voice is the voice of a person who is registered in advance, the response control unit makes a response by making a bell ring.

Therefore, the present applicant has already filed Japanese Patent Application Laid-Open No. 3-114345.
We have applied for "Caller Identification Telephone Device". By using a neural network, the recognition rate and the processing time were improved.

[0005]

However, the caller recognition telephone device proposed by the applicant of the present invention limits the target voice to the pre-learned utterance contents when recognizing the speaker. It was That is, it is premised that the caller speaks a specific word, and the usability is poor. Also, if other words are input, malfunctions may occur.

The present invention relates to a calling party recognition telephone device,
It is an object of the present invention to provide a target voice that does not need to be limited to previously learned utterances when recognizing a speaker.

[0007]

The operation of the speaker recognition unit of the present invention as set forth in claim 1 will be described. First, the neural network is trained by using the outline of the short-time spectrum obtained from the training voice. At this time, the number of data for learning is reduced by performing clustering for each speaker. At the time of recognition, an outline of the same short-time spectrum as described above is obtained from an arbitrary utterance, the sequence is input to the network, and a sequence of network output is obtained. The output vector of the obtained network suggests the speaker for each input for a short time.
One recognition is obtained by making a comprehensive judgment such as a majority vote.

The operation of the speaker recognition section of the present invention according to claim 2 will be described. First, the neural network is trained by using the outline of the short-time spectrum obtained from the training voice. At this time, the number of learning data is reduced by performing clustering for each speaker. At the time of recognition, an outline of the same short-time spectrum as described above is obtained from an arbitrary utterance, the sequence is input to the network, and a sequence of network output is obtained. The output vector of the obtained network suggests a speaker for each input for a short time, but a threshold for selecting an output vector is provided, and only a reliable output vector is selected among them. One recognition result is obtained by making a comprehensive judgment such as a sum, a product, or a majority decision for all of these.

However, the description of the "neural network" in the present invention is as follows (1) to (4).

(1) From the structure of the neural network, the neural network and the hierarchical network shown in FIG.
It can be roughly classified into two types of mutual connection networks shown in (B). The present invention may be configured by using either of both networks, but the hierarchical network is more useful because a simple learning algorithm as described later has been established.

(2) Network Structure As shown in FIG. 4, the hierarchical network has a hierarchical structure including an input layer, an intermediate layer, and an output layer. Each layer is composed of one or more units. The coupling is only forward coupling such as input layer → middle layer → output layer, and there is no coupling in each layer.

(3) Unit structure The unit is a model of a brain neuron as shown in FIG. 5, and the structure is simple. Receive input from other units,
The sum is taken and converted according to a certain rule (conversion function), and the result is output. A variable weight, which represents the strength of the bond, is attached to each of the bonds with other units.

(4) Learning (Back Propagation) Learning a network is to bring an actual output close to a target value (desired output). Generally, the conversion function and weight of each unit shown in FIG. Is learned by changing.

As a learning algorithm, for example, Rumelhart, DE, McClelland, JL and the PDP Re
search Group, PARALLEL DISTRIBUTED PROCESSING, the
Backpropagation described in MIT Press, 1986. can be used.

[0015]

In the speaker recognition method according to claim 1, by using the speaker recognition technology that does not limit the content of the utterance, it is not necessary for the caller to utter a specific word, and the caller recognition is performed. The usability of the telephone device has improved. In addition, it is now possible to respond to any words entered.

Since the representative vector of a plurality of data is used as the learning data by clustering, it is possible to reduce the number of learning data while maintaining the learning effect. As a result, the learning time of the neural network can be greatly reduced.

Furthermore, in the speaker recognition method according to the second aspect, by selecting a highly reliable output vector from among the output vectors, comprehensive judgment becomes more reliable and the recognition rate is improved.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram showing an embodiment of a calling party recognition telephone device, FIG. 2 is a schematic diagram showing a threshold value for speaker determination and an output value of a network, and FIG. 3 is a neural network. FIG. 4 is a schematic diagram showing a hierarchical neural network, and FIG. 5 is a schematic diagram showing the structure of a unit.

The caller recognition telephone device 10 recognizes whether the caller is a pre-registered speaker and enables the telephone basic circuit 11 to respond only to the registered speaker. At this time, as shown in FIG. 1, the telephone device 10 includes an automatic response unit 12 that automatically responds to an incoming call, and a speaker recognition unit that recognizes a speaker from the voice of the caller who responds to the automatic response unit 12. 1
3 and the recognition result of the speaker recognition unit 13, the telephone basic circuit 1
1 and a response control unit 14 for controlling 1). 1A is a partner telephone device.

However, the usage images of the caller identification telephone device 10 are shown in Tables 1 and 2. Table 1 is a usage image when learning the voice of the registrant, and Table 2 is a usage image when a call is received.

[0021]

[Table 1]

[0022]

[Table 2]

The speaker recognition system by the speaker recognition unit 13 will be described below. Learning text for 5 registrants and 25 non-registrants, sampling frequency 10 kHz, frame length 25.6 msec
, Fourier analysis with a frame period of 12.8 msec, 100 to 50
Obtain a sequence of 68ch (1/12 Oct.) power vectors in the 00 Hz band.

Hierarchical clustering is performed from these power vectors to obtain about 200 representative vectors for each speaker.

These representative vectors are used as the input of the neural network (input layer 68 units, the input pattern is obtained by the number of speakers × the number of representative vectors after clustering), and the corresponding output unit is activated only in the case of the registrant. To learn enough.

A sequence of power vectors is obtained in the same manner as for any utterance. This is input to the network learned by, and the sequence of output vectors {x ¹ , x ² , ... X ⁿ } x ^t = (x ^t ₁ , ..., x ^t ₅ ) n = the number of frames is obtained.

The following three methods are used for the above vector series to determine whether the input is a registered person or a non-registered person.

(1) The maximum value of Σ _t x ^t _s (s = 1 to 5) is
Registered person if it exceeds the preset threshold for speaker determination, otherwise non-registered person

(2) The maximum value of Π _t x ^t _s (s = 1 to 5) is
Registered person if it exceeds the preset threshold for speaker determination, otherwise non-registered person

(3) max {x ^t ₁ , ..., X ^t ₅ } = x ^t _s (s =
If the maximum number of 1 to 5) exceeds the preset threshold for speaker determination, it is a registered person, otherwise it is a non-registered person.

The following methods may be used instead of the above three methods. (1) Only the maximum value of Σ _t x ^t _s (s = 1 to 5) exceeds the preset first speaker determination threshold θ ₁ and the other values are preset to the second value. If it is below the speaker judgment threshold θ _{2 of} , it is a registered person, otherwise it is a non-registered person.

(2) Only the maximum value of Π _t x ^t _s (s = 1 to 5) exceeds the preset first speaker determination threshold value θ ₁ , and other values are preset. If it is below the second speaker judgment threshold value θ ₂ , the registered person; otherwise, the non-registered person

(3) max {x ^t ₁ , ..., X ^t ₅ } = x ^t _s (s
= 1 to 5) exceeds the _first threshold value θ ₁ for speaker determination set in advance and the other values have second threshold value θ 2 for speaker determination set in advance. Registered if below ₂ , otherwise non-registered

[0034] As an example of any utterance, learning short "Tomorrow will be the absence is I'm sorry because I get to Tokyo." For, "I'm home", "Hello" speaker by using the three-word of "good morning" When I did a recognition experiment,
We were able to fully recognize the 5 registered people who were used for learning and 26 non-registered people who were not used for learning.

[0035]

As described above, according to the present invention, in the caller recognition telephone device, when the speaker is recognized, it is not necessary to limit the target voice to the previously learned voice content. You can

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a caller identification telephone device.

FIG. 2 is a schematic diagram showing a speaker determination threshold value and a network output value.

FIG. 3 is a schematic diagram showing a neural network.

FIG. 4 is a schematic diagram showing a hierarchical neural network.

FIG. 5 is a schematic diagram showing a structure of a unit.

[Explanation of symbols]

 10 Caller Recognition Telephone Device 11 Telephone Basic Circuit 12 Automatic Response Unit 13 Speaker Recognition Unit 14 Response Control Unit

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Office reference number FI technical display location H04M 3/42 P

Claims (2)

[Claims] 1. A caller recognition telephone device which recognizes whether a caller is a pre-registered speaker and can respond only to a registered speaker by a telephone basic circuit. An automatic response unit, a speaker recognition unit for recognizing a speaker from the voice of a caller who replies to the automatic response unit, and a response control unit for controlling the telephone basic circuit based on the recognition result of the speaker recognition unit. The speaker recognizing unit inputs a series of vectors representing the outline of a short-time spectrum when recognizing a speaker using a neural network, and outputs a network output series of recognition results by individual outputs. 1 based on sum, product, majority vote, etc.
A caller-recognition telephone device for obtaining one recognition result, wherein the number of learning data of the network is reduced by clustering. 2. A caller recognition telephone device which recognizes whether a caller is a pre-registered speaker and can respond to the registered speaker only by a telephone basic circuit. An automatic response unit, a speaker recognition unit for recognizing a speaker from the voice of the caller who replies to the automatic response unit, and a response control unit for controlling the telephone basic circuit according to the recognition result of the speaker recognition unit. The speaker recognition unit, when recognizing a speaker using a neural network, inputs a series of vectors representing the outline of a short-time spectrum, and sets a threshold for output vector selection from a series of network outputs. With respect to the output vector selected by using, a single recognition result is obtained by summing, multiplying, majority voting, etc. of recognition results by individual outputs, and the number of learning data of the network is reduced by clustering. Caller recognition telephone apparatus according to claim the door.