KR200156360Y1 - Sound pattern loading apparatus - Google Patents

Abstract

The present invention recognizes a different voice for each channel of each voice recognition device in order to automatically dial a registered telephone number according to the input name by recognizing a unique voice pattern of each subscriber in a voice dialing service (VDS) system. The present invention relates to a voice pattern loading device for a speaker-dependent voice recognition system that allows a user to load a pattern so that multiple subscribers can receive voice services simultaneously with different voice patterns. The general voice dial service system is a speaker independent recognition system. The non-subscriber can receive the service only by knowing the subscriber's phone number and password, which makes it difficult to maintain the confidentiality of each subscriber and also causes problems with the leakage of the phone number. In this design In order to recognize the subscriber's voice as speaker-dependent so that the voice dialer service is not provided to other people, the device for loading voice recognition patterns of different subscribers into the voice recognition device is provided. The subscribers can be served at the same time, and the problem of confidentiality of the subscriber and the leakage of the telephone number is solved.

Description

Speech Pattern Loading Device for Speaker-dependent Speech Recognition System

1 is a schematic structural block diagram of a general voice recognition and answering system.

2 is a schematic diagram of a speech pattern loading apparatus for a speaker-dependent speech recognition system of the present invention.

* Explanation of symbols for main parts of the drawings

1: exchange 2: communication control unit

3: automatic response processing unit 4: voice recognition processing unit

5: voice memory unit 6: voice recognition device

7: Automatic Response Memory Unit 8: Automatic Response Device

61: Register 1 62: Register 2

63: digital signal processor A 64: digital signal processor B

65: Sram A 66: Sram B

67: SRAM B 'OR: Oagate

The present invention recognizes a unique voice pattern of each subscriber in a Voice Dial Service (VDS) system and automatically dials a registered telephone number according to a name inputted to each channel of each voice recognition device. The present invention relates to a speech pattern loading apparatus for a speaker-dependent speech recognition system for loading different speech recognition patterns so that multiple subscribers can receive voice services simultaneously with different speech patterns.

In general, the voice dialing service (VDS) enters a name to be called through the voice dialing service (VDS) system instead of manually dialing the telephone number of the party to be called by hand. It is a cutting-edge service that finds phone numbers and automatically connects them to make a call.

Each of the services provided here includes a voice dialing service that recognizes voice and automatically makes a call, a registration service for registering a called number in the name of the person to be called, and a caller number entered in the name of the call to be deleted. Delete service, a change service for changing the called number entered in the name of the called party, a search service for searching the called number entered in the called name, all names with the called number Reading service to retrieve the called party number, restricted service to set the range of callable area, password changing service to change subscriber's own password, and message when the call destination is busy or no answer Voice message service to store and deliver the voice message, you can check whether the voice message delivery result There is not such message service.

If the user wants to receive other services except the voice dialing service, the service mode button assigned to the system is switched to the service mode state, and then each service is received in the state. In addition, if the user wants to receive the service through a phone other than a registered subscriber, for example, a public telephone, the subscriber must dial the subscriber's telephone number and the password for the subscriber. At this time, the voice service phone number through the subscriber phone and the voice service phone number through another phone other than the subscriber phone are set differently, and the program cannot be used without knowing the subscriber password only in the voice service phone number for the other phone. It is.

When the subscriber enters the name of the registered telephone number by voice through the above process, the voice content is recognized and the call is automatically called when the recognition result is correct, or the name is input again.

This is a system that allows individual subscribers to make a phone call even if they remember the name without memorizing the number if they have a lot of phone numbers to remember. Anyone who needs the subscriber's phone number and password can receive the service. This is generally called a speaker independent speech recognition system.

1 is a block diagram of a general automatic answering and speech recognition system.

A communication control unit 2 which is in charge of communicating with the exchange 1; An automatic response processor (ARPC) 3 for controlling each circuit; A voice memory section 5 storing names registered from each subscriber; Voice Recognition Processor Card (VRPC) 4 which inputs a subscriber's telephone number from ARPC 3 and extracts a name registered for the telephone number from voice memory 5; )Wow; When an empty channel is occupied by the ARPC 3, the voice input processing unit VRPC 4 inputs the current input voice signal of the subscriber input from the communication control unit 2, and inputs it from the voice recognition processing unit VRPC. A Voice Recognition Unit (VRU) 6 comprising a plurality of voice recognition units for confirming whether or not the subscribers match the names registered for the numbers; An auto answer memory unit (7) for storing telephone numbers to be automatically called for names stored for each subscriber in the voice memory unit (5); The automatic response processing unit (ARPC) stores the information in the automatic response memory unit 7 stored under a name that matches the current input voice signal of the subscriber recognized by the voice recognition unit of the voice recognition unit (VRU) 6. And an automatic response unit (ARU) 8 composed of several automatic response units for input by (3) to be automatically dialed at the exchange 1.

Several voice recognition units constituting the voice recognition device (VRU) 6 have two channels, and each voice recognition unit includes five voice recognition units for each self, that is, up to six selfies, that is, up to 60 It consists of two channels. In addition, since several auto answering units forming the auto answering device (ARU) 8 each have eight channels, each self answering unit includes five auto answering units for each self. That is, up to 240 channels. This allows practically 240 subscribers to simultaneously receive voice dialing service (VDS).

The above configuration is roughly described as an independent speaker.

When the subscriber presses the voice service telephone number to receive a voice dialing service (VDS) through his / her telephone, an empty channel is occupied by the automatic answering unit 3 among the automatic answering unit of the answering machine 8, The announcement will go out. At this time, the telephone number of the subscriber who has received the current call detected by the switch 1 is input to the automatic response processing unit 3 through the communication control unit 2. Then, the automatic response processing unit 3 detects an empty channel among the voice recognition units included in the voice recognition device 6 and occupies the channel. Accordingly, the voice recognition processing unit 4 inputs the subscriber's telephone number through the automatic response processing unit 3 and loads the names stored in the voice memory unit 5 into one channel of the occupied voice recognition unit. The above operations are performed while the announcement is issued through the subscriber's telephone.

When the announcement ends, the subscriber enters the name of the registered phone number to be called automatically by voice. The voice signal is input to one channel of the voice recognition unit occupied by the communication control unit 2 and processed digitally. Since the names registered in advance for the subscriber telephone number are loaded in the occupied channel, the digitally processed subscriber current input voice signal is compared with the loaded names. Thus, if there is a name having the same pattern, it is detected by the automatic response processing unit 3 and the telephone number corresponding to the name is extracted from the automatic response memory 7. The telephone number thus extracted is analog-processed on an empty channel of the answering machine included in the answering machine 8 and automatically dialed through the switch 1.

In this speaker-independent recognition system, it is possible to provide services to a wide range of people because it is possible for other people, not subscribers, to know the subscriber's phone number and password, but it is difficult to keep confidentiality for each subscriber. It causes the problem of number leakage, and there is an inefficiency.

In view of the above, the present invention provides a device for loading voice recognition patterns of different subscribers into a voice recognition device in order to recognize the subscriber's voice as speaker dependent so that voice dialing service is not provided to other users. For the purpose of providing it.

In order to achieve the above object, a voice pattern loading device for a speaker-dependent voice recognition system of the present invention includes information on names of a telephone number to be automatically called registered by a subscriber in a voice dial service system and a subscriber's current input voice signal. In a voice pattern loading apparatus included in each channel of a voice recognition unit constituting a voice recognition device for comparing the voice recognition service, the voice dialing service is provided in speaker-dependent manner, wherein a register 1 having a unique channel number for distinguishing each channel is set. and; A register 2 for designating an address of a circuit for processing a subscriber voice currently input by an enable operation of the register 1; and a subscriber's telephone number corresponding to the subscriber's telephone number for which a voice service is currently requested by the enable operation of the register 1. Sram B and B 'with which a speech pattern is loaded; SRAM A which stores information on the feature extraction of the current input voice signal of the subscriber; A digital signal processor A for inputting a subscriber's current input voice signal and extracting a voice feature based on the information of the SRAM A; The digital signal processor A includes a digital processor B for outputting a voice signal if the voice feature extracted from the digital signal processor A matches or matches the voice patterns loaded on the SRAMs B and B '. Allow other voice patterns to be loaded. This is a device for providing a voice service required by multiple subscribers simultaneously in different voice patterns in a voice dial service system.

That is, if the channel number set in the register 1 matches the channel number instructed by the automatic response processor controlling the entire voice dial service system, the register 1 outputs an enable signal, and thus register 2 operates. At the same time as the enable operation of the register 1, subscriber voice patterns, which are repeatedly learned and stored with names registered for each subscriber, are loaded into the SRAMs B and B 'by the voice recognition processor. These operations are performed with the subscriber's telephone number recognized by the exchange when the subscriber requests the voice dial service and the initial announcement leaves.

In this way, a device for loading a unique voice pattern for each subscriber is provided so that the voice dial service can be provided speaker-dependent.

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

FIG. 2 is a block diagram illustrating a speech pattern loading apparatus for a speaker-dependent speech recognition system of the present invention. The general speech recognition and automatic response system of FIG.

The present invention is included in each channel constituting the voice recognition unit constituting the voice recognition device 6 of FIG.

The configuration is as follows.

A register 1 (61) for setting a unique channel number for distinguishing each channel and outputting an enable signal in accordance with whether the channel number specified by the control is matched to the automatic response processing unit (ARPC) 3; A register 2 (62) for designating an address of a circuit for processing subscriber voice currently input by an enable operation of the register 1 (61); An SRAM B and B '(66) 67 in which a voice pattern of a subscriber corresponding to the subscriber's telephone number requesting the current voice service is loaded by the enable operation of the register 1 (61); An SRAM A 65 which stores information on feature extraction of the subscriber's current input voice signal; A digital signal processor A (63) which inputs a subscriber's current input voice signal and extracts a voice feature by the information of the SRAM A (65); The voice signal extracted from the digital signal processor A 63 matches the voice patterns loaded on the SRAMs B and B '66 and 67, and if there is a match, the voice signal corresponding to the voice signal is automatically answered. And a digital signal processor B 64 for outputting to the (ARPC) 3.

The register 1 (61) is a channel numbering stage (PN0 to PN2) in which the order of the channels included in each voice recognition unit is determined, and a card arranged in order for the voice recognition units configured in each Self. The channel numbers set in the number stage CN0 to CN3, the data bus D designated by the automatic response processing unit ARPC 3, and the channel numbers preset in the stages PN0 to PN2 and CN0 to CN2 are If it is matched, open the corresponding channel to enable the start of operation ( ) And the write signal by the automatic response processor (ARPC) ) And chip selection signal Is composed of a clock terminal (CLK) and ground terminal (G) for inputting by OR logic.

Here, the channel number stages PN0 to PN2 represent the sequence numbers of the channels constituting the voice recognition units, and the card number stages CN0 to CN3 represent the sequence numbers of the voice recognition units and other circuit cards constituted in each of the shelves. Indicates.

And, the enable end ( ) Is an operable state when the signal of the signal is '0' and the write signal ( ) Indicates that information on the channel number is input from the ARPC 3, and the chip selection signal Is a signal that distinguishes Register 1 (61) and Register 2 (62) and prevents two registers 61 and 62 from operating simultaneously.

The register 2 (62) is operated when the register 1 (61) is enabled. The memory board designates the addresses of the digital signal processing units A and B (63) 64 which input and process the subscriber's current input voice signal. The voice patterns from the address stages MBA0 to MBA4, the digital signal processing section selection stage DSPSL for causing the digital signal processing units A and B 6-2 to operate in sequence, and the voice patterns from the speech recognition processing section 4 are the SRAM B and Reset stage for resetting digital signal processing units A and B (63) 64 when downloaded to B '(66) (67) And the test stage TST which informs the digital signal processing units A and B 63 and 64 when it is to be used for the test, and the write signal by the automatic response processing unit ARPC 3 ) And chip selection signal Is composed of a clock terminal (CLK) and ground terminal (G) for inputting by OR logic.

The digital signal processing unit selection stage DSPSL is divided into two and outputs a signal for selecting the digital signal processing units A and B 63 and 64 which differ in their operation, and the digital signal processing unit A 63 is set to '0'. ', And outputs' 1' to the digital signal processor B64.

Each stage of the registers 1, 2 (61) and 62 as described above can be expressed as follows.

Here, since the bits indicating the numbers of the channels included in the voice recognition unit are 3 bits, the number of channels expandable to the voice recognition unit is possible up to eight.

That is, the voice pattern loading apparatus of the present invention loads the subscriber's own voice pattern into each channel of the voice recognition unit to determine the subscriber's current input voice signal, so that multiple subscribers can receive services simultaneously with different voice patterns. do.

It will be described with reference to FIG. 1 of the present invention speech pattern loading device configured as described above.

When a subscriber presses a service voice telephone number to receive a voice dialing service, an empty channel is occupied by an automatic answering unit (ARPC) of the answering machine of the answering machine 8, and the announcement is sent out there. At this time, the telephone number of the subscriber who has received the current phone call detected by the switch 1 is input to the automatic response processing unit (ARPC) 3 through the communication control unit 2. Then, the ARPC 3 receives the channel number of one of the empty channels so as to occupy one of the empty channels among the voice recognition units included in the VRU 6. VRU) is sent to each voice pattern loading device of 6). The register 1 61 of each channel inputted by the automatic response processing unit (ARPC) 3 confirms whether it matches its own channel number set in the register 1 61. If the channel numbers match, register 1 (61) outputs an enable signal, and register 2 (62) operates accordingly, and outputs an address to each of the digital signal processing units A and B (63) 64. At the same time as the register 1 61 is enabled, the voice pattern of the subscriber with respect to the names registered for the subscriber telephone number stored in the voice recognition memory section 5 is inputted by the voice recognition processing apparatus (VRPC) 4. It is loaded into SRAM B and B '65, 66. At this time, the register 2 62 resets the digital signal processing units A and B 63 and 64.

These operations are performed while the announcement is out via the telephone. After the announcement, the voice signal is input by the subscriber. The voice signal is input to the digital signal processor A 63 in the voice pattern loading device of the occupied channel directly through the switch 1 and the communication control unit 2. . This voice signal is extracted by the voice feature extraction method stored in the SRAM A (65). Then, the digital signal processing unit B64 searches for whether the extracted voice feature matches the voice pattern of the subscriber loaded on the SRAMs B and B '66 and 67. At this time, if there is a match, the signal is sent to the automatic response processing unit (ARPC) 3, and the automatic response processing unit (ARPC) 3 extracts the telephone number registered in the subscriber from the automatic pressure memory unit 7; Send to answering machine (ARU) 8. The answering machine in the answering machine (ARU) 8 finds a phone number registered in the voice signal recognized by the answering machine (ARPC) 3 among the phone numbers registered for the subscriber, and communicates with the communication control unit 2. ) To be dialed automatically at the exchange (1).

As described above, a voice pattern loading device configured to load different voice patterns for each channel is configured so that a plurality of subscribers may be simultaneously provided with different voice patterns. In addition, since the service is provided according to the unique voice characteristics of each subscriber, the confidentiality of the subscriber is guaranteed.

As described above, according to the present invention, by providing a voice pattern loading device for receiving an automatic dialing service only by the subscriber's voice, multiple subscribers can simultaneously receive the service even with different voice patterns, The problem of confidentiality and leakage of telephone numbers is solved.

Claims (1)

In the voice dialing service (VDS) system, information on names of telephone numbers to be called automatically registered by a subscriber and each channel of a voice recognition unit constituting a voice recognition device (VRU) for comparing the subscriber's current input voice signal A voice pattern loading apparatus for providing a voice dial service to a speaker dependent device, comprising: a register 1 (61) having a unique channel number for distinguishing each channel; A register 2 (62) for designating an address of a circuit for processing subscriber voice currently input by an enable operation of the register 1 (61); An SRAM B and B '(66) 67 in which a voice pattern of a subscriber corresponding to the subscriber's telephone number requesting the current voice service is loaded by the enable operation of the register 1 (61); An SRAM A 65 which stores information on feature extraction of the subscriber's current input voice signal; A digital signal processor A (63) which inputs a subscriber's current input voice signal and extracts a voice feature by the information of the SRAM A (65); Search for whether the voice feature extracted by the digital signal processor A 63 matches the voice pattern loaded in the SRAM B and B '66 and 67, and if there is a match, the digital signal outputting the voice signal accordingly. Speech pattern loading device for a speaker-dependent speech recognition system characterized in that it comprises a signal processor (B) 64 so that different voice patterns are loaded into each channel.