JPH03162165A - Automatic answering telephone set - Google Patents

Abstract

PURPOSE:To constitute an automatic answering telephone set sot that only a necessary main point can be known quickly in a printed state by adding functions for recognizing an input voice, converting to a character image and recording, to an automatic answering telephone set with a facsimile. CONSTITUTION:When a voice recognition recording function is designated by a setting switch 10, a recognition recording control circuit 8 executes the control of recognition of an input voice, conversion to a character image and recording in accordance with a second output of a whole control circuit 2. A character converting circuit 9 converts an input voice to a character-string in accordance with the control of a recognition and recording control circuit 8, sends out character image data to a receiving information recording part 7, and the receiving information recording part 7 executes printing of a character graphic in accordance with an output from the character converting circuit 9. In such a way, only an important main point of the contents of a message can be confirmed exactly and quickly in a printed state.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to an answering machine.

[Conventional technology]

A conventional answering machine will be explained using an answering machine with facsimile as an example.

FIG. 3 is a block diagram showing the structure of a conventional facsimile answering machine.

In FIG. 3, the incoming/outgoing call detection circuit 1 detects whether the facsimile answering machine is in the receiving state or the outgoing state based on the call signal from the public line or the operation of the hook, and then sends it to the overall control circuit &'820. Send control signals. The overall control circuit 20 receives a control signal from the incoming/outgoing call detection circuit 1, and controls the entire telephone for sending and receiving calls and facsimile answering machine. When recording an answering machine using this facsimile answering machine, the answering machine recording control circuit 3 is connected to the first
Controls the answering machine recording by receiving the output. Answering machine recording section 4
Under the control of the answering machine recording control circuit 3, the voice message from the caller is stored or reproduced as voice. Magnetic tape or semiconductor memory is used for recording. Next, when transmitting/receiving text/graphic information using this facsimile answering machine, the fax control circuit 5 receives the second output from the overall control circuit 20 and causes the fax transmitting/receiving circuit 6 to transmit/receive text/graphic information. . The received information recording section 70 prints out the received character and graphic information according to the output from the facsimile transmitting/receiving circuit 6.

As explained above, conventional answering machines have means for recording received voice messages on a storage medium such as a magnetic tape or semiconductor memory, and reproducing them as voice.

[Problem to be solved by the invention]

With conventional facsimile answering machines, in order to check the contents of a voice message from a caller, the entire recording must be played back from the beginning, and it takes approximately the same amount of time as the recording time to check the contents. This was extremely inconvenient when you wanted to quickly learn only the essential points. Furthermore, since the contents of voice messages cannot be hard-copied and converted into files, there is a drawback that it is difficult to organize and save important messages.

[Means to solve the problem]

The answering machine of the present invention is a type of answering machine that detects a call signal, receives a call, sends out a predetermined voice message and input prompting signal to make the caller utter a desired message, and records the message. means for storing speech characteristic information extracted from predetermined types of speech signals representing different phonemes or phoneme sequences, divided into a plurality of groups; and a predetermined order of specific groups from the plurality of groups. means for transmitting an audio message instructing to utter only voices belonging to the selected group; means for extracting characteristics of the audio signal input in response to the transmitting means; and means for identifying a phoneme or phoneme sequence by comparing the phoneme or phoneme sequence with pre-extracted feature information belonging to the selected group; and means for converting the identified phoneme or phoneme sequence into a corresponding audio signal and transmitting it. means for transmitting a message requesting confirmation as to whether or not the converted and transmitted voice signal has the same phoneme or phoneme sequence as the uttered voice; means for recognizing a response to the confirmation request; and means for recognizing the recognition result. , and means for recording the corresponding characters of the phoneme or phoneme sequence in a predetermined format. [Example] Next, an example of the present invention will be explained using the drawings. FIG. 1 is a block diagram showing the basic configuration of a facsimile answering machine according to an embodiment of the present invention. The facsimile answering machine according to the embodiment of the present invention has the function of recognizing input voice, converting it into a character image, and recording it in addition to the conventional facsimile answering machine that uses sequential access memory such as magnetic tape as a storage medium for the answering machine recording function. is added. Semiconductor memory is used as the voice data memory for storing the input prompt messages necessary for the above-mentioned input voice recognition, conversion to character images, and recording operations, as it is convenient for repeated operations to confirm the voice recognition results described later. It uses random access memory such as

In FIG. 1, an incoming/outgoing call detection circuit 1 detects whether a facsimile answering machine is in an incoming or outgoing state based on a call signal from a public telephone line or a hook operation.

A setting switch 10 specifies the function selection of answering machine recording or voice recognition recording. The overall control circuit 2 controls the transmission and reception of the telephone as a single unit and the entire facsimile answering machine according to the output of the incoming call detection circuit 1 and the state of the setting switch 10. The answering machine recording control circuit 3 receives the first output from the overall control circuit 2, controls the answering machine recording, and causes the answering machine recording section 4 to store the voice message from the caller or reproduce it as voice. The recognition recording control circuit 8 performs recognition of input speech, conversion to character images, and recording control in accordance with the second output of the overall control circuit 2. The character conversion circuit 9 converts the input voice into a character string under the control of the recognition recording control circuit 8, and sends character image data to the received information recording section 7. The fax control circuit 5 controls fax transmission and reception according to the third output of the overall control circuit 2, and causes the fax transmission and reception circuit 6 to transmit and receive character and graphic information.

The reception information recording unit 7 prints characters and figures according to the outputs from the fax transmission/reception circuit 6 and the character conversion circuit 9. FIG. 2 shows the recognition recording control circuit 8 of the embodiment shown in FIG.
, is a diagram showing in detail the structure of the character conversion circuit 9 and the received information recording section 7.

In FIG. 2, an audio data memory 82 stores audio data that is classified into groups and to be reproduced in a predetermined order, such as input prompt messages, personal names, numbers, etc., for example. As mentioned above, semiconductor memory is used as the storage medium. The audio reproducer 83 reproduces audio according to the data read from the audio data memory 82. The voice analyzer 84 analyzes the characteristics of the input voice emitted as a response from the caller according to the way the voice reproducer 83 outputs it. The standard pattern memory 85 stores, in a predetermined order, features obtained by analyzing standard sounds that are classified into groups in advance, such as personal names and numbers. The pattern matching circuit 86 successively compares the output of the speech analyzer 84 and the contents of the standard pattern memory 85 and outputs the storage address of the optimal standard pattern.

The aforementioned audio data memory 82, audio reproducer 83, audio analyzer 84, standard pattern memory 85 and pattern matching circuit 86 are controlled by a processor 81 for program control.

The processor 81 is activated by a signal input from the overall control circuit 2. The character conversion circuit 9 is composed of a character image conversion circuit 91 and a fixed character image memory 92. The character image conversion circuit 91 converts character image data of phonemes or phoneme sequences within a predetermined group based on the storage address of the optimal standard pattern output from the pattern matching circuit 86 under the control of the processor 81. Fixed character image memory 92 stores character image data of predetermined fixed character portions. The received information recording section 7 is composed of a line image buffer 71 and a printing device 72. line image buffer 7
1 temporarily stores character image data for each scanning line output from the character image conversion circuit 91 and fixed character image memory 92 of the character conversion circuit 9. Printing device 7
2 prints the scanning signal output from the line image buffer 71. Next, the operation of the first embodiment will be explained. Note that the answering machine recording function and facsimile transmission/reception function of a conventional facsimile answering machine are well-known techniques, and therefore their explanation will be omitted in the description of this embodiment. In the embodiment of the present invention shown in FIG. 1, when making a call from a facsimile answering machine, for example, when the handset is lifted, the incoming call detection circuit 1 detects the movement of the handset hook, and the state changes to the call making state. A control signal for this is sent to the overall control circuit 2. When a call is received, the incoming call detection circuit detects the call signal from the public line, connects the line, determines whether the signal sent through the connected line is a voice signal or a fax signal, and makes the determination. A control signal based on the result is sent to the overall control circuit 2. When the incoming call detection circuit 1 detects a fax signal, it sends a command to receive the fax to the overall control circuit 2. The overall control circuit 2 sends a control signal to the fax control circuit 5 to start receiving the fax. When the incoming signal is other than a fax signal, the overall control circuit 2
determines the overall reception mode from the output of the incoming call detection circuit 1 and the setting state of the setting switch 10 (for example, the setting of the answering machine recording function or the voice recognition reception function), and sends the control signals necessary for the operation of the reception mode. , and sent to the answering machine recording control circuit 3 or the recognition recording control circuit 8.

Hereinafter, detailed operations when the setting switch 10 is set to the voice recognition reception function will be explained using FIG.

The voice recognition reception function is selected by the reception preparation operation described above, and a control signal is input from the overall control circuit 2 to the recognition recording control circuit 8, and voice recognition and conversion of the recognition result into a character image are performed.

The overall control circuit 2 resets the processor 81 and causes it to execute a predetermined program. The processor 81 accesses the audio data memory 82 at the beginning of the program and reads the audio data of the first input prompting message of the first group. The read audio data is input to the audio reproducer 83 and is reproduced into an audio signal in the audible band.

The first message in the first group includes the name of the recipient and a message announcing their absence, as well as an announcement that a record will be taken using voice recognition and a limited recognition range, similar to those in conventional answering machines. In addition, a message or the like instructing the contents of the utterance within a predetermined group is transmitted as reproduced audio. For example, ``We will record your name as text using voice recognition.Please speak your name clearly.''1 or ``We will record your name as ``18'' using voice recognition.Please separate your name into letters and speak clearly. ” Nato. In this example, the first group described above is limited to processing related to the sender's personal name, and in the subsequent recognition operation, recognition is performed within the group of personal names.

After the first message of the first group is finished, the processor 81 stops accessing the audio data memory 82, stops the operation of the audio reproducer 83, and activates the audio analyzer 84 for a predetermined period of time. , to analyze the voice input from the public line.

The voice analyzer 84 analyzes the input voice as a response from the caller and extracts features. The features to be extracted include conventionally known LPG parameters, LSP parameters, and PAR.
Any feature such as COR coefficient or formant may be used. Furthermore, as an analysis method, either an analog or digital method may be used for the audio input through the line. Here, for the sake of simplicity, a case will be described in which a digital method is used.

The speech analyzer 84 analyzes the input speech as a response from the caller and sends the result to a first input of the pattern matching circuit 86 .

On the other hand, the processor 81 sequentially reads the feature data of the recognition candidates belonging to the first group from the standard pattern memory 85 and sends them to the second input of the pattern matching circuit 86 . The pattern matching circuit 86 performs the audio '! f+
IEl, 8 4 output data and standard pattern memory 8
The processor 81 temporarily stores the output from the pattern matching circuit 86 and stores the audio data memory 82 based on the output. The personal name voice data corresponding to the recognition result is read out from the name voice data storage area in the first group, inputted to the voice reproducer 83, converted into an audio signal in the audible band, and sent out.

Furthermore, the processor 81 accesses the voice data memory 82, reads out data from the storage area for the confirmation prompting message, inputs it to the voice reproducer 83, converts it into a voice signal, and sends it out as in the case of the person's name. The two voice messages mentioned above are combined, and the caller hears, for example, ``This is Mr. Sato.If you agree, please answer rYes.''If it is incorrect, please reply rNo1.''The caller hears this as a voice. After sending the above two messages, the processor 81 operates the voice analyzer 84 again for a certain period of time to analyze the voice of the caller's response in the same manner as in the case of personal name recognition described above.
However, this time, the data in the standard pattern memory 85 to be input to the pattern matching circuit 86 is read from the area for storing confirmation response data and sent.

The processor 81 monitors the output of the pattern matching circuit 86, and if the confirmation result is "yes", it moves on to recognition processing for the next group. If “No”, count the number of times voices within the same group could not be recognized. If the count value exceeds a predetermined value, a status is sent to the recognition record control circuit 8. In the embodiment of the present invention, a forced termination function for repeated recognition errors is not particularly provided, but if there are many errors in the recognition results and it is difficult to continue, the recognition recording control circuit 8 sends a forced termination request signal to the overall control circuit 2. issue, terminate the voice recognition recording,
It is also possible to set the general reception mode again to the conventional answering machine recording function. If forced termination is not used, the processor 81 returns to the process immediately after the reset.

When the person's name is correctly recognized, the processor 81 inputs the previously stored recognition result of the person's name (the storage address of the optimal standard pattern) and the group number currently being processed to the character image conversion circuit 91. Character image conversion circuit 91
generates the corresponding character image data based on these two inputs. Conversion to character image data can be achieved using existing ROM, etc. The character image data output from the character image conversion circuit 91 is temporarily stored in the line image buffer 71 as character image data for each scanning line. Similarly, from the fixed character image memory 92, based on the group number, character image data according to a predetermined format, such as "sender" in the mode of recognizing and converting a person's name into characters.
Character image data such as is read out. Output from character image conversion circuit 91 and fixed character image memory 9
The outputs from 2 are combined and stored in a predetermined arrangement in a line image buffer 71. When the call ends and the caller hangs up, the incoming call detection circuit 1 detects disconnection of the line and sends a call termination signal to the overall control circuit 2. The overall control circuit 2 stops the operation of the character conversion circuit 9, and finally causes the printing device 72 to print the character image data stored in the line image buffer 71. An example of printing the character images accumulated through the above processing is shown below.

In the above printing example, the part marked with *l on the right side of the character string is the part where a fixed character image stored in advance is printed. The part marked with *2 is the part where the converted character image was printed after voice recognition.

Also, the parts marked with * are parts in which a clock function is built into the overall control circuit 2 in advance, and its output is converted into characters and printed. In addition, for audio such as sentences that are difficult to recognize, such as business items, options are provided in advance in the input prompt message, the number or symbol of the option is uttered, and the character string corresponding to the number or symbol is printed. It is also possible to improve recognition processing. In the example above, the message ``Please select from the following for your business.''

1. Please call me. 2 I'll call you again.

3. Please call ・・・・・・・・・. 4 Send a fax. Please call me to confirm. Now, please take your number. ” to convert and print the character image shown in the example above. In the embodiment of the present invention described above, the magnetic tape used in the answering machine recording section 4 of the answering machine recording function is used as the storage medium of the audio data memory 82, since it is convenient for controlling repetitive operations when there is a speech recognition error. Although we used a semiconductor memory that can be randomly accessed in addition to a sequential access memory such as the above, it is of course possible to use a sequential access memory as well. Furthermore, if a random access memory is used in the answering machine recording section 4, the memory for answering machine recording and the voice data memory for voice recognition can be stored in the same memory. [Effects of the Invention] According to the present invention, by voice recognition and transcribing the recognition results, when confirming a message left by voice, all the recorded contents must be sequentially played back. This eliminates the disadvantage of requiring approximately the same playback time as the original message, and makes it possible to clearly and quickly confirm only the important points of the message content in printed form.
Furthermore, by predetermining the print format, it becomes possible to create a hard copy file of the message content.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the basic structure of a facsimile answering machine according to an embodiment of the present invention, and FIG. 2 is a block diagram showing the structure of a recognition record control circuit, a character conversion circuit, and a received information recording section according to an embodiment of the present invention. FIG. 3 is a block diagram showing the basic rules of a conventional facsimile answering machine. DESCRIPTION OF SYMBOLS 1... Incoming call detection circuit, 2.20... Overall control circuit, 3... Answering machine recording control circuit, 4... Answering machine recording section, 5... Fax control circuit, 6... Fax transmission/reception Circuit, 7, 70... Reception information recording section, 8...
Recognition recording control circuit, 9...Character conversion circuit, 10...
Setting switch, 71... Line image controller, 72... Printing device, 81... Processor sensor, 82
...Audio data memory, 83...Audio reproducer, 84
...Speech analyzer, 85...Standard pattern memory, 8
6...1<turn processing circuit, 91...Character image conversion circuit, 92...Fixed character image memory.

Claims (1)

[Claims] An answering machine of the type that detects a call signal, receives a call, sends a predetermined voice message and input prompting signal to make the caller utter a desired message, and records the message, means for dividing and storing speech characteristic information extracted from a predetermined type of speech signal representing a phoneme or a phoneme sequence into a plurality of groups, and storing specific groups from the plurality of groups in a predetermined order. means for selecting; means for transmitting a voice message instructing to utter only voices belonging to the selected group; means for extracting features of the voice signal input in response to the transmitting means; and the extraction result. means for identifying a phoneme or phoneme sequence by comparing it with previously extracted feature information belonging to the selected group; means for converting the identified phoneme or phoneme sequence into a corresponding audio signal and transmitting the same; means for transmitting a message requesting confirmation as to whether or not the converted and transmitted voice signal has the same phoneme or phoneme sequence as the uttered voice; means for recognizing a response to the confirmation request; and means for transmitting a message based on the recognition result. 1. An answering machine, characterized in that it has means for recording the corresponding character of the phoneme or phoneme string in a predetermined format.