JPH0797277B2 - Computer and communication device for unspecified speakers - Google Patents

Description

TECHNICAL FIELD The present invention relates to communication with machines, and more particularly to voice communication of commands with electronic computers.

BACKGROUND ART A great deal of interest has been directed to and researched on human voice communication with electronic computers. The advantages, applications, and general equipment of voice communication with electronic calculators are: "Toss Your Ke
yboards and Just Tell Your Computer What to Do "
(AEConrad, Research & Development, January 1984 issue,
Vol. 26, No. 1, pp. 86-89).

As pointed out herein, most research in this area has focused on speaker-specific or speaker-dependent devices and on language-recognition devices that are speaker-independent devices. It was The former can only recognize the words spoken by an individual,
The latter can recognize the same words spoken by different people. A speaker-specific device can recognize (or learn) more words at the same time than a speaker-independent device, thus allowing more control of the function. Both devices rely on creating a template of spoken words and matching received words to this template.

A great deal of effort has been put into this area, and Marley's US patent
There are many patents and articles proven in a search of the technical literature in the prior art description of 4,284,846. This prior art patent teaches an apparatus for analyzing and comparing words by comparing certain waveform characteristics with pre-stored ratios.

Other patents disclosing language recognizers include Sahoe U.S. Pat. Nos. 4,286,114 and 4,319,221; BHAn U.S. Pat.
U.S. Pat.Nos. 4,319,085 and 4,3, 4,292,470, Welch et al.
36,421, Kellett U.S. Pat.No. 4,343,969, Pirz U.S. Pat.No. 4,349,700, Taniguchi et al U.S. Pat.
109, Hitchock US Pat. No. 4,388,495, Duifhuis et al. US Pat. No. 4,384,335, and Rothschild et al. US Pat. No. 4,399,732.

These devices are extremely complex and expensive, or otherwise have very limited capabilities. For example, on a typical speaker-independent device, about 10 words (eg 10 different numbers)
There is a device that recognizes 100 to 200 words, which is one digit larger in the speaker identification device.

SUMMARY OF THE INVENTION The present invention comprises the devices disclosed in the above referenced patents,
The difference is that the complexity is relatively low. That is, the present invention is typically needed in conversation in order to economically implement a speaker-independent device capable of recognizing a large number of different voice commands. Alternatively, there is a physical advantage that a human voice in a range far exceeding the range normally used can be used. An apparatus in accordance with the teachings of the present invention senses and recognizes differences in sound that are treated as encoded signals to perform a predefined function.

According to the characteristics of the present invention, this difference in sound is a sound within a scale such as a normal diatonic scale, and, for example, the difference between the user's central C-sound and the subsequently pronounced D-sound. Means are provided for setting the tonality (key) for each user by comparing the sounds. The device of the present invention uses and recognizes pitches as commands or inputs.

The device finds application in many applications, including computer input and control in environments where other input devices are impractical.
That is, the present invention is effective when the user has to use his hands in a dark room (for example, in an electron microscope room) or during other operations such as a manufacturing process. This device is convenient for people with disabilities, especially those who cannot easily operate conventional computer terminals.

The recognition device of the present invention together with its advantages can best be understood by the following description with reference to the accompanying drawings. In the drawings, similar components are provided with similar reference numbers.

Description of the Preferred Embodiments The principles of the present invention may be applied in various ways with hardware and / or software, some examples of which are described below.

The device according to the invention as a whole, as shown in FIG.
Designated by reference numeral 10. The device 10 includes a transducer 12, such as a microphone, which collects a sound wave as shown by waveform 14. The above waveform has a fundamental period T for the rate of the fundamental frequency (pitch) when a human is singing a Ha sound. Received speech waveform 14 to simplify the electronic circuit system
The electric analog signal of is converted into a pulse wave train 16 having a period T by the waveform shaping circuit 18.

Instead of the converter 12, an auxiliary electrical signal input 20 may be provided. This may be, for example, a remotely activated telephone input, or a voice generator.

The wave train 16 is received by a period measuring circuit 22, its period is measured and digitized, and the information is sent to an input / output interface 24 which interfaces with a numerical computer 26.

In the overall operation of the device 10, the device 10 uses the microphone 12
Alternatively, a voice signal is received at the input 20 and a preprogrammed routine in the computer is executed in response to a signal selected from the received signals.

In order to understand the principle of operation of the device 10 of the present invention, an application example will be described with reference to a normal diatonic scale. Therefore, it is understood that the present invention can be applied to other scales. The modern diatonic scale has a pitch that is independent of the frequency of a particular note (but when one frequency is set, the frequency of another note is determined by it), and is partly represented by the table below. obtain.

Any scale pitch MI ′ 2.5000 RE ′ 2.2500 DO ′ 2.0000 TI 1.8750 LA 1.6667 SO 1.5000 FA 1.3333 MI 1.2500 RE 1.1250 Reference scale DO 1.0000 TI, 0.93750 LA, 0.83333 DO, 0.75000 FA, 0.66667 MI, 0.62500 RE, 0.56250 DO, 0.50000 TI, 0.46875 Most people are familiar with this pitch, and even children can immediately sing Do, Re, Mi, Fa, So, La, Si, Do (in this case, unconsciously select the reference frequency, Other notes are related to the reference frequency in the above pitch). The present invention selects a pitch as a means for receiving information and processes it.

Therefore, referring to the flow chart of FIG. 5, the start command may be the reception of the audio signal (h) detected in step A. In step B, this signal is measured to determine if it has a period of sufficient duration (prevention of false triggers). If a reference signal is present and stored (eg, do), they are compared, and when a pitch of 1.125 is calculated in step C, a fetch / execute subrun is executed in step D and is pre-recorded. Call and execute a subroutine for pitch 1.125. Finally, the system is reset in step E, ready to receive a second signal (eg, fa) and respond to it in the same general manner.

The reference signal is set in the same process in order to avoid the system becoming speaker specific and to be usable by anyone with a minimum of musical ability. At start-up, the user only needs to blow the microphone into the microphone for a short period of time. The system treats the first received voice signal as a reference signal setting and treats the second and subsequent signals as command signals.

FIG. 2 shows a preferred embodiment of the waveform shaping circuit 18 and the interconnections to the microphone 12 and auxiliary input 20. Although electrical specific values for the components used are shown in the figure, many other values can of course be used, as is well known to those skilled in the art. However, the values and connections of the circuit elements in Figure 2 were good in the prototype device.

Specifically, the microphone 12 is connected between the reactive impedance 27 to the circuit connection point 28 and the chassis ground point. Connection point 28 is connected to auxiliary input 20 via resistor 32 to chassis ground via capacitance 34.

When a signal is received at signal input 12 or 20, input 12 or 20
Sends the received signal to a low frequency amplifier 36 including an operational amplifier 38. The negative input (pin 6) of the amplifier 38 is connected to the connection point 28, and the positive input (pin 5) is connected to the bias voltage (+ 5V) via the resistor 40 and the chassis ground point via the parallel connection of the capacitor 42 and the resistor 44. Connected to. Pin 4 of the operational amplifier 38 is connected to the positive bias source (12V) via the current limiting resistor 45 and to the chassis ground point via the capacitor 46. A part of the output of the operational amplifier 38 is fed back to the negative input through the parallel connection of the resistor 47 and the capacitor 48.

The main output of low frequency amplifier 36 is a low frequency filter including resistor 51.
Supplied to 50. One end of the resistor 51 is connected to the output of the amplifier 36, and the other end is connected to the ground point via the capacitor 52 and the circuit connection point 54 via the resistor 53. The output of the low frequency filter 50 is connected to a connection point 54, and further connected to a comparator 55 and a maximum voltage follower circuit 56 with attenuation. The signal is (a) connected via a current isolation diode 57 to the main positive signal input (pin 12) of an operational amplifier 58 of a comparator 55 whose input is grounded via a resistor 59, and (b) a maximum voltage follower circuit 56. Connected to. The output of circuit 56 is the major negative signal input to the input pin (pin 13) of comparator 55.

Circuit 56 preferably includes operational amplifier 60, and amplifier 60
Has its primary positive input (pin 3) connected to node 54, its output (pin 1) directly returned to its negative input (pin 2), and partly connected to the negative signal input of amplifier 58 via isolation diode 61. To be done. The output of the resistor 60 is grounded via the resistor 63 and the capacitor 64. The discharge time constants of the resistor 63 and the capacitor 64 connected in parallel always affect the inverting input of the comparator 55 more than the non-inverting input, except for the maximum effective value of each cycle of the signal appearing at the connection point 54. Is the value.

Waveform 16, which is the output of comparator 55, has a set of resistors 65 connected in series from the output of operational amplifier 58 (pin 14) to ground.
And 66 connection points. This output is the period measurement circuit
Sent to 22 input STs.

FIG. 3 shows a preferred embodiment of the interconnection of the period measuring circuit 22, the input / output interface circuit 24 and the computer 26. The particular computer 26 is preferably an Apple II ⁺ , and here is shown the particular interconnect to that computer.

The pulse train 16 is an operational amplifier that acts as a Schmitt trigger.
Coupled to shift register 72 via 70. Register 72
The output of is coupled to the counter 78 via the period measurement gate 75 of the inverter 74. The counter 78 is connected to the computer 26 via the buffer 80.

The output of the calculator 26 is R / W, AO, Al. It is derived from the output and bias source (5V) and timing pulses and is supplied as shown in FIG. Gates 81 and 82 are
It is for sending a reset command via the signal line 83.

The function of the circuit of FIGS. 2-3 and the calculator 26 will become more apparent with reference to FIG. Figure 4 shows waveforms 14 and 16
And the input of the shift register 72 at CP, the timing pulse sent from the calculator 26 to the period measuring circuit 75, and the output of the shift register 72 corresponding to the beginning and end of the measured period T.

The operation will be described. The circuit of FIG. 2 and FIG. 3 first has 1 at the input point ST (FIGS. 2 and 3) to the Schmitt trigger 70.
The input signal 14 is shaped into a train of pulses 16 to obtain one pulse per cycle. The computer 26 resets the shift register 72 and the counter (via the inverter 86 via the signal line 83) and sets up the circuit for the next period measurement.

In the CP of the shift register 72, the shaped pulse train 16
When the first rising edge of appears, output Q0 changes from a low (voltage) level to a high level. The counter 78 begins measuring the current period of the input wave train.

In the CP of the shift register 72, the second of the shaped pulse train 16
When the rising edge of arrives, output Q1 goes from low to high. The counter 78 stops counting. At the same time, the input of the highest digit of the high byte buffer 90 of buffer 80 is set from low to high, indicating that the period measurement is complete.

Calculator 26 reads high byte buffer 90. If the most significant digit is high, it is ignored and the computer 26 evaluates the least significant 7 bits as a 15-bit binary number, Q8-Q14. (If the highest digit is low, it indicates that the period measurement is not yet completed.) The computer 26 also reads the low byte buffer 92 and reads the lowest 8 bits of the 15-bit binary number, Q0 to Q7. And The calculator 26 sets the size of this 15-bit binary number as the size of the period just measured.

The operation of device 10 can be better understood from the flow chart of FIG. In the figure, the start sequence A proceeds to the measurement circuit reset (gates 81-82 and associated leads), and further at B1 it is determined whether a periodic signal has been received. If not, the system waits for the signal. If received, take readings at B3 and at B4
In step C, it is determined whether the J-th (any number, for example, 30th) pulse having the same pulse period can be used for the average value calculation.

The flow chart of FIG. 6 is for a single pitch message. This is for separate reference and signal wave train inputs, there is always a break between the continuous wave trains. Block I is the step for detecting these breaks.

If the output of block I is the first of the detected audio signals, that signal is treated as a reference signal and stored. Once the second and subsequent signals have been identified, they are compared to the reference signal and the pitch calculated. The calculated pitch is checked for a match, and if a match is found, the associated subroutine is executed and the program is reset.

In a practical example, the computer terminal could display the following menu.

I AM AT YOUR SERVICE.'SING'YOUR CHOICE (DO DO) FOR (LIST PROGRAMME IN MEMORY) (DO RE) FOR (DISPLAY PATTERN'HO ') (DO ME) FOR (TEXT MODE DISPLAY) (DO FA) FOR ( FLASH MODE DISPLAY) (DO SO) FOR (PLAY RUNNING TONES) (DO LA) FOR (ACTIVATE EXTERNAL DRIVE TO CATALOG P
ROGRAMMES ON DISK) (DO TE) FOR (DISPLAY'TE ') (DO DO') FOR (ACTIVATE EXTERNAL DRIVE TO SAVE TH
IS PROGRAMME ON DISK, AND EXECUTE ANOTHER PROGRAMME
ON DISK, AND RETURN) To start the computer, the user only has to say a request command. The following is an optimal listing for the program of FIG.

By modifying the listing line 93 as follows, the device 10 operates with a new reference signal for each command.

93 W = φ The above operation allows the same or different speakers to freely change the reference signal for each command.

FIG. 7 shows another flow chart 100 for the apparatus of the present invention. This figure is a program for N pitch message.
That is, it is coding for multiple sounds. For example, here the doremi and the drafer are different signals.

Below is a listing suitable for running this program.

FIG. 8 shows still another flow chart of the voice program. By inputting a series of voices, the speaker inputs a voice program including a desired voice command sequence to be executed thereafter (for example,
Effectively define a process).

Following a flow chart of FIG. 8, a suitable listing for executing this program is shown below.

FIG. 9 is another subroutine II, which can be replaced with the block I in the flowcharts of FIGS. 6 to 8, and if the block I in FIGS. 6 to 8 is replaced, input with slur, that is, slur Allows processing of the wave train. This behavior is advantageous for the speaker, as it is easier to make slurred (concatenated) speech than individual speech.

The following listings are suitable for implementing the Block II program.

2420 PRINT CHR $ (7): REM BEEP FOR NEXT WAVE TRAI
N 2460 for PS = 1 TO 500: NEXT: REM PAUSE This can be replaced with, for example, lines 2420 to 2485 of the program listing for the flowchart of FIG. The pause in FIG. 9 is the wave train at that time (flow charts in FIGS. 6 to 8).
Should be long enough to avoid being mistaken for subsequent readings, for example, for the next wave train, and the user Care should be taken not to continue to generate for more than a pause.

FIG. 10 shows another alternative subroutine III. Block I
The II modifies the device 10 so that the user can actually expand the frequency range by generating and holding the wave train output for longer than the time normally required. That is, it suffices to generate and hold the data over the pause time. When the device detects a wave train that lasts longer than a predetermined duration, it obtains a substantial pitch before message identification (ie, before interpretation).
The conversion factor is used to modify the above data.

The symbol m shown in block III of FIG. 10 is a conversion coefficient, which can take a value in an arbitrary range. Two specific values 0.5 and 2 are particularly useful when a human is the speaker. When the coefficient is m = 0.5, the device listening section transposes to a high octave, and when m = 2, it shifts to a low octave. (That is, Dore has a pitch of 1.125. If m = 5 and RE is held, the pitch is 2.5.
That is, measure DO-RE '). Repeated transposition is performed when the speaker or singer continues to continue the wave train. That is, the speaker frequency range is substantially expanded. Further, the voice input allows the user to operate in a frequency range that is easy for the user to use, and yet realizes a large number of pitches as if the speaker has a very wide frequency range. Therefore, more different word signals than before can be obtained using only a few notes.

The following listings are suitable for the block III program of FIG.

2420 PRINT CHR $ (7): REM BEEP FOR NEXT WAVE TRAI
N 2464 FOR PS = 1 TO 1000: NEXT: REM PAUSE 2466 A = PEEK (49348): REM RESET CIRCUIT AND TEST F
OR SILENCE 2470 FOR PS = 1 TO 100: NEXT: REM BRIEF PAUSE 2472 REM "WAVE" TRAIN STILL DETECTED? "2475 L = PEEK (49345): H = PEEK (49346) 2478 IF L = 0 THEN GO TO 2495: REM NO WAVE DETECTED 2480 AVE (W) = AVE (W) / 2: REM MULTIPLYING FACTOR
= 1/2 2485 GO TO 2420 It is clear from the above description that the present invention teaches a new device for electronic computer communication. The device uses the concept of pitch-based communication, which essentially extends the number of voice commands and the ease of voice recognition.

This device can be realized economically and effectively, and its operation can be easily learned and used effectively. It provides a large number of recognized words (tones) without being user-specific and is economical to manufacture and use.

Although several embodiments of the apparatus of the present invention have been shown and described, it will be obvious to those skilled in the art that changes and modifications to the apparatus 10,000 or variations thereof described herein may be made without departing from the teachings of the present invention. Is. Therefore, the scope of the present invention is limited only to the matters described in the claims.

In summary, the present invention relates to an apparatus for communicating to a machine that uses pitch as a code for communicating preselected instructions or inputs to a computer. Since this device sets the pitch continuously, it can use the voice of the person who is generating the note (dore etc.), converts the received note or voice into a pulse train of the same fundamental period, and Electrical circuits and programs can be used to store and calculate pitches. When a specific pitch is received, a device containing a numerical microcomputer such as the Apple II + TM executes a pre-stored subroutine, followed by reconfiguring the circuit for the reception of the second pitch. Is possible.

[Brief description of drawings]

FIG. 1 is a block diagram of a speech recognition apparatus constructed in accordance with the teachings of the present invention, showing waveforms at each point, FIG. 2 is a circuit diagram of a waveform shaping circuit section of the apparatus of FIG. 1, and FIG. Is a circuit diagram of the period measuring circuit and other parts of the apparatus of FIG. 1, FIG. 4 is a waveform diagram showing a set of waveforms for understanding the operation of the circuit of FIGS. 2 and 3, FIG. 5 is a flow chart for illustrating the overall operation of the apparatus shown in FIGS. 1 to 3, and FIGS. 6 through 8 are flow charts for showing the operation of another embodiment of the present apparatus. 9 and 10 are sub-routine flow charts that can replace the portions of the flow charts of FIG. 6, FIG. 7, or FIG. Explanation of symbols of main parts 12 …… Converter, 18 …… Wave shaping circuit 20 …… Auxiliary input terminal, 22 …… Period measurement circuit 24 …… Input / output interface 26 …… Electronic computer, 36 …… Low frequency amplifier 50 ...... Low-frequency filter, 55 …… Comparator 72 …… Shift register, 75 …… Period measurement gate 78 …… 15-bit counter 80 …… 8-bit counter

Claims (15)

[Claims] 1. An apparatus for communicating using a voice of an unspecified speaker with a computer in which a series of subroutines corresponding to various pitches are stored in advance, wherein the apparatus transmits an input voice having a repeating period. Conversion means for converting into a signal representing the repetition period, pitch calculation means coupled to the conversion means for calculating the pitch of the received voice and reference voice signal, and the conversion means and the pitch calculation means in response to the conversion means, An electronic computer and a communication device for an unspecified speaker, comprising: a pre-stored subroutine corresponding to a specific pitch calculated by the pitch calculating means, and a selection running means for running the subroutine. 2. A device according to claim 1, wherein the device is capable of responding to a reference voice signal set by a voice input to the converting means in the order of voice of a command. A computer and a communication device for unspecified speakers. 3. An apparatus according to claim 2, wherein the apparatus is capable of responding to musical tones within a musical scale, and a communication apparatus for a computer and an unspecified speaker. 4. A computer according to claim 3, wherein said scale is a modern scale having a pitch including 1.000, 1.250, 1.500 and 2.000 within a permissible error range. Communication device for a specific speaker. 5. A voice activated computer controller, said device comprising: converting means for converting a voice into a pulse train of said voice's repetitive period; and coupled to said converting means to measure the period of said pulse train,
A comparison means for comparing this with a reference cycle, a pitch calculation means for calculating a pitch between the cycle of the pulse train and the reference cycle, and a response to all the above means for responding to a selected and calculated specific pitch value. And a means for causing an electronic computer to execute various routines. 6. A human voice recognition device, said device having means for responding to a command as a voice corresponding to a specific pitch, for recognizing a plurality of specific different commands, the same number of commands being different. A human voice recognition device characterized in that it corresponds to a pitch. 7. A device according to claim 6, wherein said one of said voices expands or contracts the detected pitch when one of said voices repeats continuously over a preselected time. , A human voice recognition device which operates to generate different pitches. 8. An apparatus for recognizing a voice of an unspecified speaker, the apparatus responding to a voice input of a person and generating a pulse train having the same repetition period as that of the voice input, and a waveform shaping circuit of the waveform shaping circuit. A period measuring and comparing circuit for receiving and measuring the pulse train output, the period measuring and comparing circuit comprising means for comparing the period of the pulse train with a reference period, and means for calculating its pitch, An apparatus for recognizing a voice of an unspecified speaker, further comprising means for responding only to a specific pitch and not responding to other pitches. 9. A method for communicating a human voice with an electronic computer through an interface circuit, the method comprising: providing a command as a voice to the interface circuit; calculating a pitch of a received voice with respect to a reference value of the voice. And a step of selecting and executing, in the computer, a subroutine protocol recorded in advance corresponding to a calculated specific pitch, in the computer, the method of communicating a human voice. 10. A method according to claim 9, wherein the voice is a musical tone in a musical scale. 11. A computer according to claim 10, wherein the scale is a modern scale having a pitch including 1.000, 1.250, 1.500 and 2.000 within a permissible error range. Communication method of human voice. 12. The method according to claim 11, wherein the method first converts the reference voice signal into a signal indicating a voice cycle of the reference voice signal, and then converts the voices of the commands in that order. Then, a method of communicating between a computer and a human voice, comprising the step of converting into a signal showing the repetition period and then calculating the pitch. 13. A method according to claim 9, wherein the method comprises the steps of: detecting the point in time when the voice is repeated over a continuous period; and expanding or contracting the detected repeated pitch. And a step of generating different pitches, the method of communicating a human voice with an electronic computer. 14. A method according to claim 9, wherein the method includes the step of converting each voice into a pulse train having a specific period corresponding to the voice. Voice communication method. 15. The method according to claim 12, wherein the converting step includes converting the reference voice signal into a pulse train having a specific period corresponding to a voice of the reference voice signal, and ,
Converting each voice into a pulse train corresponding to the voice,
And a step of calculating the pitch, the communication method between a computer and a human voice.