TWI766375B - System of bank counter based on artifical intelligence - Google Patents

Abstract

A system of bank counter based on artificial intelligence comprises a voice recognition device, a processing device, and a multi-function printer. The voice recognition device obtains a voiceprint characteristic, transaction item information, and transaction parameter information according to at least one voice input. The processing device is connected to the speech recognition device, determines that the voiceprint characteristic conforms to pre-stored identity information, determines a form format based on the transaction item information, and generates a form file based on the form format, the pre-stored identity information, and the transaction parameter information. The multi-function printer is connected to the processing device, and receives and prints the form file.

Description

AI-based bank counter system

The present invention relates to a bank counter system, in particular to a bank counter system based on artificial intelligence.

Under the existing banking operation mode, customers are usually required to manually fill in the transaction form, and then submit it to the counter clerk, who will input the transaction items, amount and other transaction information filled in the form. Such bank over-the-counter transaction services not only take a long time to operate, but also cannot avoid typing errors during the input process of the bank staff, resulting in wrong transaction information being entered into the back-end system.

In view of the above, the present invention provides a bank counter system based on artificial intelligence.

An artificial intelligence-based bank counter system according to an embodiment of the present invention includes a voice recognition device, a processing device, and a transaction machine. The voice recognition device obtains voiceprint features, transaction item information and transaction parameter information according to at least one voice input. The processing device is connected to the voice recognition device, judges that the voiceprint feature conforms to the pre-stored identity information, determines the form format according to the transaction item information, and generates a form file according to the form format, the pre-stored identity information and the transaction parameter information. The MFP is connected to the processing device, and receives and prints the form file.

With the above structure, the bank counter system based on artificial intelligence disclosed in this case can link printing operations through the voice recognition mechanism, replacing the traditional manual filling in forms, saving customers time, reducing the error rate of clerks and improving service efficiency. Effect.

The above description of the present disclosure and the following description of the embodiments are used to demonstrate and explain the spirit and principle of the present invention, and provide further explanation of the scope of the patent application of the present invention.

The detailed features and advantages of the present invention are described in detail below in the embodiments, and the content is sufficient to enable any person skilled in the relevant art to understand the technical content of the present invention and implement it accordingly, and according to the content disclosed in this specification, the scope of the patent application and the drawings , any person skilled in the related art can easily understand the related objects and advantages of the present invention. The following examples further illustrate the viewpoints of the present invention in detail, but do not limit the scope of the present invention in any viewpoint.

Please refer to FIG. 1 , which is a functional block diagram of an artificial intelligence-based bank counter system according to an embodiment of the present invention. As shown in FIG. 1 , the bank counter system 1 based on artificial intelligence includes a speech recognition device 11 , a processing device 13 and a transaction machine 15 , wherein the processing device 13 can be connected to the speech recognition device 11 and the transaction machine 15 by wireless or wire. The voice recognition device 11 can obtain the user's voiceprint feature, transaction item information and transaction parameter information according to at least one voice input of a user (eg, a customer who wants to perform financial transactions). Among them, the user's voiceprint feature represents the user's identity and can be used for subsequent identification procedures; the transaction item information includes the financial transaction items the user wants to perform, such as loan, remittance, deposit, withdrawal, etc.; transaction parameter information is It can include the amount corresponding to the financial transaction item the user wants to carry out, such as 2,000 yuan, 100,000 yuan, etc. In addition, the transaction parameter information can also include the object of the financial transaction item the user wants to carry out, such as a remittance account.

Further, the speech recognition device 11 may include a speech input interface 111 , a vocabulary database 113 and a processor 115 . The voice input interface 111 is, for example, a receiver such as a microphone, and is used for receiving voice input to be recognized, wherein the so-called voice input is, for example, a sound signal sent by a user. The vocabulary database 113 at least stores a plurality of pre-stored transaction item vocabulary and a plurality of pre-stored number vocabulary, and further, the vocabulary database 113 can store a plurality of pre-stored object vocabulary. The vocabulary database 113 may be composed of one or more non-volatile storage media such as flash memory, ROM, magnetic memory, and the like. For example, the above-mentioned pre-stored transaction item vocabulary and pre-stored number vocabulary can be respectively stored in two independent non-volatile storage media. For another example, the above-mentioned pre-stored transaction item vocabulary and pre-stored number vocabulary can be respectively stored in different magnetic areas in the same non-volatile storage medium, or in different storage blocks defined by computer programs. The vocabulary database 113 may be located in a remote memory (eg, a cloud drive) relative to the voice input interface 111 and the processor 115 , or may be located in one or both of the voice input interface 111 and the processor 115 . memory in the same enclosure. The processor 115 is, for example, a central processing unit, a microcontroller, a programmable logic controller, etc., and can be connected to the voice input interface 111 and the vocabulary database 113 by wireless or wired means. The processor 115 of the voice recognition device 11 can perform voiceprint capture and voice recognition on the voice input received by the voice input interface 111, so as to obtain voiceprint features, transaction item information and transaction parameter information from the voice input, and use the obtained information It is sent to the processing device 13, and the detailed identification process will be described later. In addition, in addition to the above-mentioned components, the voice recognition device 11 may further include an output interface (such as a speaker, a screen, etc.) for outputting a voice input request by means of voice or screen display.

The processing device 13 is, for example, a computer host, and can transmit signals to and from the speech recognition device 11 in a wireless or wired manner. The processing device 13 can determine the form format according to the transaction item information obtained by the voice recognition device 11 when it is determined that the voiceprint feature obtained by the voice recognition device 11 conforms to the pre-stored identity information, and according to the determined form format and the pre-stored identity information and transaction parameter information to generate a form file. Further, the processing device 13 may include a voiceprint database and a form format database. The voiceprint database stores multiple pre-stored identity information and the voiceprint features corresponding to each pre-stored identity information. For example, the voiceprint database can store the names of multiple users as the pre-stored identity information, the voiceprints of these users features, as well as the corresponding relationship between each user's name and voiceprint features. Wherein, the voiceprint feature is, for example, the waveform feature of the user's voice signal, including a plurality of formant frequency bandwidth, pitch contour, syllable duration, long-term average power spectrum parameter, and long-term average pitch parameter. The form format database can store multiple form formats, and each form format corresponds to different financial transaction items. For example, the internal memory of the processing device 13 may store form formats suitable for loans, remittances, deposits and withdrawals, respectively. The above-mentioned voiceprint database and form-format database may be composed of one or more non-volatile storage media such as flash memory, read-only memory, magnetic memory, and the like. For example, the voiceprint database and the form-format database are respectively two non-volatile storage media. For another example, the voiceprint database and the form-format database are respectively different magnetic areas in the same non-volatile storage medium, or different storage areas defined by computer programs.

The processing device 13 may further include one or more internal processors (eg, a central processing unit, a microcontroller, a programmable logic controller, etc.) to process two programs of voiceprint recognition and form format determination. The processing device 13 may first execute the voiceprint recognition program to determine whether the voiceprint feature obtained by the voice recognition device 11 matches the pre-stored identity information, and execute the form format determination program when it is determined that it matches. Further, the internal processor of the processing device 13 can compare the voiceprint features obtained by the speech recognition device 11 with the voiceprint features pre-stored in the voiceprint database. When a matching voiceprint feature is found, it is determined that the voiceprint feature has matching pre-stored identity information, and then the form format determination procedure is executed. In particular, as mentioned above, the voiceprint feature may include multiple parameters, and when the processing device 13 determines that the captured voiceprint feature is more than 80% identical to the voiceprint feature in a pre-stored identity information, it determines The voiceprint feature matches the pre-stored identity information. The form format determination procedure includes selecting a form format corresponding to the financial transaction item included in the transaction item information obtained by the speech recognition device 11 from a variety of form formats in the form format database, and pre-stores the identity information (eg user name) and The text corresponding to the transaction parameter information (such as amount, transaction object) is filled in the file with the form format, and then the file is used as a form file, or after format conversion, it is used as a form file.

The MFP 15 is, for example, a multifunctional MFP, which can be connected to the processing device 13 in a wireless or wired manner, so as to receive form files from the processing device 13 and print the form files. Further, after generating the form file, the processing device 13 then transmits the form file to the MFP 15, and also transmits a print command so that the MFP 15 executes the printing operation after receiving the form file. Specifically, the document obtained by printing the form by the transaction machine 15 includes the form corresponding to the form format and the text corresponding to the pre-stored identity information and transaction parameter information. In one embodiment, in addition to the above operations, the MFP 15 can perform scanning according to a newly added command, wherein the newly added instruction can be triggered by the user through a key of the MFP 15 . For example, when there is a new form format, the bank clerk can press the newly added button on the transaction machine 15 to scan the document with the new form format to generate a new form format scan file. In this embodiment, the processing device 13 has an optical character recognition function, and the processing device 13 can perform optical character recognition on the scanned file in the new form format, and use the result of the optical character recognition as a new financial transaction item , and the scanned file of the new form format is used as a new form format, and the two and the corresponding relationship between the two are stored in the form format database.

In addition to the above devices, the bank counter system proposed in this case may further include other optional devices. Please refer to FIG. 2 , which is a functional block diagram of an artificial intelligence-based bank counter system according to another embodiment of the present invention. As shown in FIG. 2 , the bank counter system 1 ′ includes a voice recognition device 11 , a processing device 13 , a transaction machine 15 , a verification information capture device 17 and a display device 19 . The connection relationship and cooperative operation among the recognition device 11 , the processing device 13 and the multi-functional machine 15 are the same as those of the speech recognition device 11 , the processing device 13 and the multi-functional machine 15 in the previous embodiment of FIG. 1 , and will not be repeated here. .

The verification information capturing device 17 is an optional device, which can be connected to the processing device 13 by wireless or wired. The verification information capture device 17 is, for example, a signature pad with a speaker or a screen or a biometric (fingerprint, iris, etc.) capture device, which can be controlled by the processing device 13 to output a verification request to an external (such as a user) and capture the pending verification request. Verify information (e.g. signature, biometrics, etc.). In addition to the operations described in the embodiment of FIG. 1 , the processing device 13 can also store a high-risk type list, recording one or more high-risk transaction item information (such as loans) or transaction parameter information (such as more than 1 million transactions). amount or unknown transaction object). In addition, the processing device 13 can also store signatures, biometric features or other verification information corresponding to each of the plurality of pre-stored identity information, and the corresponding relationship between the pre-stored identity information and the verification information, in addition to storing the voiceprint features.

The processing device 13 can determine whether the transaction item information or transaction parameter information belongs to the high-risk type before generating the form file, and when the determination result is yes, the processing device 13 will control the verification information capture device 17 to display the method through voice or screen Output the verification request to the user, and when the verification information capture device 17 captures the information to be verified provided by the user and returns it, it then determines whether the information to be verified conforms to the pre-stored identity information, and generates a form only when the determination result is yes file. In this way, the bank counter system 1' can improve the accuracy of the identity verification and the security of the system through the multi-stage identity verification procedure of voiceprint recognition and verification information confirmation.

In another embodiment, in addition to being connected to the processing device 13 , the verification information capturing device 17 is also connected to the MFP 15 through a wireless or wired manner. In this embodiment, after the form file is generated and before the form file is sent to the MFP 15, the processing device 13 can control the verification information capture device 17 to output the verification request to the user by means of voice or screen display. When the capturing device 17 captures and returns the information to be verified provided by the user, it then determines whether the information to be verified conforms to the pre-stored identity information, and only transmits the form file to the transaction machine 15 when the determination result is yes. In this way, in addition to improving the accuracy of identity verification and the security of the system through the multi-stage identity verification process, the bank counter system 1 ′ can be used before the confidential documents (ie, the above-mentioned form files) are transmitted from one device to another device. Confirming the necessity of transmission can reduce unnecessary transmission, and thus reduce the risk of robbery during transmission, compared to the authentication procedure that is performed after the confidential document is output.

The display device 19 is also an optional device, and can be connected to the voice recognition device 11 by wireless or wired. The display device 19 is, for example, a liquid crystal display, and is used for displaying one or more of the voiceprint feature, transaction item information and transaction parameter information obtained by the speech recognition device 11 . In this embodiment, the voice recognition device 11 can display voiceprint features, transaction item information and transaction parameter information through the display device 19 in addition to the operations described in the embodiment of FIG. The pattern features, transaction item information and transaction parameter information are sent to the processing device 13 . The confirmation instruction is, for example, a voice input indicating confirmation, which may originate from the user. Alternatively, the speech recognition device 11 may include other input interfaces such as a touch screen, buttons, etc., to receive the user's touch or press action and regard it as a confirmation command. Through the setting and operation of the display device 19, the accuracy of the identity verification can be improved. In addition, the confirmation command can also be used as a reference for correcting the internal algorithm of the speech recognition device 11, thereby increasing the accuracy of speech recognition.

Please refer to FIG. 1 to FIG. 3 together. FIG. 3 is a flowchart of an operation method of a bank counter system according to an embodiment of the present invention. The operation method shown in Fig. 3 is applicable to the bank counter systems 1 and 1' described in the previous embodiments of Figs. 1 and 2 . In step S1, the bank counter system 1 or 1' obtains the voiceprint feature, transaction item information and transaction parameter information through the voice recognition device 11 according to at least one voice input of a user (such as a customer who wants to perform a financial transaction). In step S3, the bank counter system 1 or 1' determines by the processing device 13 that the voiceprint feature conforms to the pre-stored identity information, determines the form format according to the transaction item information, and generates a form file according to the form format, the pre-stored identity information and the transaction parameter information . In step S5, the bank counter system 1 or 1' receives and prints the form file from the processing device 13 via the transaction machine 15 for the user to confirm. After confirming the information on the document printed out according to the form file, the user can sign and seal the document, and hand the document to the bank clerk for transaction. With the above operation method, the customer only needs to confirm the transaction information on the form and then sign and seal, no need to handwrite the form, which can improve convenience and save transaction time, and the clerk does not need to choose which form to use, and the system automatically determines the transaction items. After printing, it can reduce the error rate of clerk registration and improve work efficiency.

In particular, in one embodiment, after the voice recognition device 11 transmits the acquired voiceprint features, transaction item information and transaction parameter information to the processing device 13, the voice input, voiceprint features, and transaction item information are deleted. and transaction parameter information records, and the MFP 15 also deletes the form file after printing is completed. In this way, the bank counter system 1 or 1' can reduce confidential information (such as the above-mentioned records of voice input, voiceprint features, transaction item information and transaction parameter information) stored in the peripheral devices in the system (such as the above-mentioned voice recognition device 11 and time of the transaction machine 15), thereby reducing the risk of the theft of confidential information and improving the security of the system.

The detailed execution content of the above-mentioned step S3 is the same as the operation of the processing device 13 described in the previous embodiment, and the detailed execution content of the step S5 is also the same as the operation of the MFP 15 described in the previous embodiment, and will not be repeated here. Step S1 is further described, ie, the step of acquiring the voiceprint feature, transaction item information and transaction parameter information according to at least one voice input performed by the voice recognition device 11 . The voice input can be the user's voice signal, and the voice recognition device 11 can capture the waveform characteristics of the voice signal, such as formant frequency bandwidth, pitch contour, syllable duration, long-term average power spectrum parameters, long-term average tone parameters Several of them are used as voiceprint features. In particular, the voice recognition device 11 may perform the extraction of voiceprint features for each received voice input.

As for the acquisition of transaction item information and transaction parameter information, please refer to FIGS. 1 to 4 , wherein FIG. 4 is a flowchart of speech recognition in the operation method of the bank counter system according to an embodiment of the present invention. In step S11 , the processor 115 of the speech recognition device 11 takes the sound signal received from the speech input interface 111 as a speech input to be recognized, and uses the speech input to be recognized as the execution object of the subsequent recognition program. In step S13, the processor 115 of the speech recognition device 11 executes a recognition program on the execution object, the recognition program includes judging whether the execution object includes the target transaction item vocabulary and the target number vocabulary, wherein the target transaction item vocabulary is the vocabulary in the vocabulary database. One of the plurality of pre-stored transaction item vocabulary, and the target number of words is one of the plurality of pre-stored number of words in the vocabulary database. Further, the processor 115 converts the sound signal input as the voice to be recognized into a pinyin character string, uses the multiple pre-stored transaction item vocabulary in the vocabulary database as the judgment target one by one, and judges whether the pinyin character string has a judgment target, When the judgment result is yes, the current judgment target is used as the vocabulary of the target transaction item. The judgment of the target number words is the same as the judgment of the above-mentioned target transaction item words, so it is not repeated here.

When the processor 115 determines that the execution object includes the target transaction item vocabulary and the target number vocabulary, it will use the target transaction item vocabulary as transaction item information and the target number vocabulary as transaction parameter information (step S15 ), and complete the transaction item information and transaction parameters. acquisition of information. On the contrary, when the processor 115 determines that the execution object does not include one or both of the target transaction item vocabulary and the target number vocabulary, it will output a voice input request to receive another voice input to be recognized, and the other voice to be recognized The input and the previously judged voice input to be recognized are used as the execution object (step S17 ), and then the recognition procedure described above is executed (step S13 ). And so on, until the transaction item information and transaction parameter information are obtained from all the voice inputs to be recognized.

The content of the voice input request will be further explained. In one embodiment, when the speech recognition device 11 determines that the execution object does not contain the target transaction item vocabulary, the speech recognition device 11 generates a voice input request including the keyword corresponding to the transaction item information; and when the execution object does not include the target number When the vocabulary is selected, the voice recognition device 11 generates a voice input request including keywords corresponding to the transaction parameter information. For example, when the execution object includes the target transaction item vocabulary of "loan" but does not include the target number of words, the content of the voice input request may include "please input the desired transaction amount". For another example, when the execution object contains the target number vocabulary of "1 million" but does not contain the target transaction item vocabulary, the content of the voice input request may include "Please input the desired transaction item". The above-mentioned voice input request can be output in the form of voice or screen display through the output interface of the voice recognition device 11 , such as a speaker or a screen.

In another embodiment, the internal memory of the processing device 13 stores the frequently used transaction items and the frequently used numbers corresponding to each pre-stored identity information, wherein the frequently used transaction items are among a plurality of pre-stored transaction item vocabularies in the speech recognition device 11 . At least one of , and the common number is at least one of a plurality of pre-stored number words in the speech recognition device 11 . When the voice recognition device 11 determines that the execution object does not contain the target transaction item vocabulary, the voice recognition device 11 can first provide the acquired voiceprint feature to the processing device 13 to determine the pre-stored identity information corresponding to the voiceprint feature, and the processing device 13 then determines the pre-stored identity information according to the pre-stored identity information. The identity information selects its corresponding common transaction item to generate a voice input request, and sends it back to the voice recognition device 11 for output; when the voice recognition device 11 determines that the execution object does not contain the target number of words, the voice recognition device 11 can first obtain the The voiceprint feature is provided to the processing device 13 to determine the pre-stored identity information corresponding to the voiceprint feature, and the processing device 13 selects the corresponding common number according to the pre-stored identity information to generate a voice input request, and returns it to the voice recognition device 11 for output. For example, the processing device 13 may store the common numbers of a user as “1000 yuan” and “2000 yuan” according to the previous usage record. When the speech recognition device 11 determines that the execution object includes the target transaction item vocabulary of “withdrawal” but not When the target number of words is included and the processing device 13 determines that the voiceprint recognition belongs to the above-mentioned user, the processing device 13 can generate a voice input request including "1000 yuan" and "2000 yuan", and the voice recognition device 11 can display the voice or screen way to output the options of "1000 yuan" and "2000 yuan" for users to choose. The voice recognition device 11 can predict the content of the voice input to be recognized in the next input according to the voice input requirements generated by the common transaction items and the common numbers, thereby reducing the computational load of the next recognition procedure and improving the execution speed.

With the above structure, the bank counter system based on artificial intelligence disclosed in this case can link printing operations through the voice recognition mechanism, replacing the traditional manual filling in forms, saving customers time, reducing the error rate of clerks and improving service efficiency. Effect.

Although the present invention is disclosed in the foregoing embodiments, it is not intended to limit the present invention. Changes and modifications made without departing from the spirit and scope of the present invention belong to the scope of patent protection of the present invention. For the protection scope defined by the present invention, please refer to the attached patent application scope.

1. 1’ Bank counter system 11 Voice recognition device 111 Voice input interface 113 Vocabulary database 115 processors 13 Processing device 15 business machine 17 Authentication Information Capture Device 19 Display device

FIG. 1 is a functional block diagram of an artificial intelligence-based bank counter system according to an embodiment of the present invention. FIG. 2 is a functional block diagram of an artificial intelligence-based bank counter system according to another embodiment of the present invention. FIG. 3 is a flowchart of an operation method of a bank counter system according to an embodiment of the present invention. FIG. 4 is a flowchart of speech recognition in the operation method of the bank counter system according to an embodiment of the present invention.

1 Bank counter system 11 Voice recognition device 111 Voice input interface 113 Vocabulary database 115 processors 13 Processing device 15 business machine

Claims (7)

A bank counter system based on artificial intelligence, comprising: a voice recognition device, which obtains a voiceprint feature, a transaction item information and a transaction parameter information according to at least one voice input; a processing device, connected to the voice recognition device, to determine the The voiceprint feature corresponds to a pre-stored identity information, a form format is determined according to the transaction item information, and a form file is generated according to the form format, the pre-stored identity information and the transaction parameter information; an MFP is connected to the processing device and receives and print the form file; wherein the speech recognition device includes a vocabulary database, the vocabulary database stores a plurality of pre-stored transaction item vocabulary and a plurality of pre-stored number words, and the speech recognition device executes the acquisition according to the at least one voice input The transaction item information and the transaction parameter information include: receiving a voice input to be recognized, executing a recognition program, and using the voice input to be recognized as an execution object of the recognition program, wherein the recognition program includes determining whether the execution object includes the a target transaction item vocabulary in the pre-stored transaction item vocabulary and a target number vocabulary in the pre-stored transaction item vocabulary; When the execution object does not contain the target transaction item vocabulary or does not contain the target number of vocabulary, output a voice input request to receive another to-be-recognized voice input, and the to-be-recognized voice input and the other to-be-recognized voice input are used as the executing the object and executing the identification procedure; and when the execution object includes the target transaction item vocabulary and the target number vocabulary, use the target transaction item vocabulary as the transaction item information and the target number vocabulary as the transaction parameter information; and a display device connected to the voice recognition device, wherein the voice recognition device further displays the acquired voiceprint feature, the transaction item information and the transaction parameter information through the display device; wherein the processing device does not include When the target transaction item vocabulary is selected, at least one of the pre-stored transaction item vocabulary is selected according to the pre-stored identity information, and the voice input request is generated according to the at least one of the pre-stored transaction item vocabulary, and the processing device is in When the execution object does not contain the target number of words, select at least one of the pre-stored number of words according to the pre-stored identity information, and generate the voice input request according to the at least one of the pre-stored number of words. The bank counter system according to claim 1, wherein when the execution object does not contain the target transaction item vocabulary, the voice input request contains corresponding In the keyword of the transaction item information, and when the execution object does not contain the target number of words, the voice input request contains the keyword corresponding to the transaction parameter information. The bank counter system according to claim 1, wherein the voice recognition device further deletes the voice input, the voiceprint feature, The transaction item information and the transaction parameter information, and the MFP deletes the form file after printing. The bank counter system according to claim 1, wherein the transaction machine further performs scanning according to a new instruction to generate a new form format scan file, the processing device performs optical character recognition on the new form format scan file, and The result of the optical character recognition, the scan file of the new form format and the corresponding relationship between the two are stored. The bank counter system according to claim 1, further comprising a verification information capture device connected to the processing device, wherein the processing device determines that the transaction item information or the transaction parameter information belongs to a high level before generating the form file. When the risk type is selected, the verification information capturing device is controlled to output a verification request, receive a pending verification information corresponding to the verification requirement, and generate the form file after judging that the pending verification information conforms to the pre-stored identity information. The bank counter system according to claim 1, further comprising a verification information capture device connected to the processing device and the MFP, wherein the processing device further controls the verification information capture device to output a Verifying a request, receiving a to-be-verified information corresponding to the verification request, and after judging that the to-be-verified information conforms to the pre-stored identity information, the form file is sent to the MFP. The bank counter system according to claim 1, wherein the voice recognition device further transmits the voiceprint feature, the transaction item information and the transaction parameter information to the processing device through the display device after receiving a confirmation command.

Images