EP0453511A1 - Procede et appareil de traitement de la parole - Google Patents

Procede et appareil de traitement de la parole

Info

Publication number
EP0453511A1
EP0453511A1 EP90902869A EP90902869A EP0453511A1 EP 0453511 A1 EP0453511 A1 EP 0453511A1 EP 90902869 A EP90902869 A EP 90902869A EP 90902869 A EP90902869 A EP 90902869A EP 0453511 A1 EP0453511 A1 EP 0453511A1
Authority
EP
European Patent Office
Prior art keywords
telephone
speech
signal
stored
digital signal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP90902869A
Other languages
German (de)
English (en)
Other versions
EP0453511A4 (en
Inventor
Peter J. Schmuckal
Mitchell S. Phillips
Francis P. Keiper, Iii
James C. Sprout
Ronald H. Fried
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Origin Technology Inc
Original Assignee
Origin Technology Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US07/294,168 external-priority patent/US5007081A/en
Application filed by Origin Technology Inc filed Critical Origin Technology Inc
Publication of EP0453511A1 publication Critical patent/EP0453511A1/fr
Publication of EP0453511A4 publication Critical patent/EP0453511A4/en
Withdrawn legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04M—TELEPHONIC COMMUNICATION
    • H04M13/00—Party-line systems
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/08—Speech classification or search
    • G10L15/12—Speech classification or search using dynamic programming techniques, e.g. dynamic time warping [DTW]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04M—TELEPHONIC COMMUNICATION
    • H04M1/00—Substation equipment, e.g. for use by subscribers
    • H04M1/26—Devices for calling a subscriber
    • H04M1/27—Devices whereby a plurality of signals may be stored simultaneously
    • H04M1/271—Devices whereby a plurality of signals may be stored simultaneously controlled by voice recognition

Definitions

  • the present invention relates to a speech processing apparatus having particular application in a speech activated telephone.
  • Speech recognition apparatuses are well known in the art.
  • a speech recognition apparatus can be used to activate a number of task including a telephone.
  • the algorithm used in speech recognition is complex and has required the use of a dedicated signal processor.
  • the use of a dedicated signal processor has increased the cost of the apparatus.
  • DT basic dynamic time warping
  • Warping Algorithm for Isolated Word Recognition by K. Paliwa, A. Agarwal and S.S. Sinha, IEEE Int'l Conf. On Acoust. Speech and Si ⁇ . Proc.. Vol. ICASSP- 2, pp. ' 1259-61, May 1982.
  • this algorithm attempts to solve the problem of window skewing, it is also subject to error.
  • Speech activated phones are also well known in the art. However, they have not provided a mechanism by which questionable choices of the particular speech pattern to a stored speech pattern can be resolved. Thus, they are prone to error and cannot be resolved by user input. These and other considerations have not permitted a speech activated telephone to use an inexpensive general processor wherein other novel features may be implemented, without a great deal of cost.
  • a single line telephone is connected to a single pair of physical wires labeled tip and ring. Since communication must be effected in both directions through the tip and ring lines, a balance transformer has been used to isolate the tip and ring lines from the transmit and receive lines internal within the telephone. Such a balance transformer is expensive and is bulky.
  • Prior art telephones have provided for the monitoring of the telephone line to which it is attached; However, the typical monitoring functions have been limited to ringing, hold, and busy. The telephone apparatus has not been able to display the status of whether or not the telephone is connected at all to the line without placing the telephone off hook to determine if the telephone is connected to the line.
  • a speech recognition apparatus and method therefor uses a modified clipped auto correlation function wherein a first difference of the speech signal is obtained prior to applying the clipped auto correlation function to produce speech patterns, which are stored.
  • the apparatus also uses a constant width band dynamic time warping processing algorithm and adaptive linear prune line algorithm to match the input speech pattern to the stored speech pattern.
  • the present invention also relates to a speech activated telephone that uses a modified clipped auto correlation function to process speech signals to produce speech patterns and to store the speech patterns.
  • the telephone also uses a constant width band dynamic time warping processing algorithm and adaptive linear prune line algorithm to match the input speech pattern with the stored speech pattern. Further, the telephone provides for user input on questionable matches.
  • the telephone has a user interface menu for entering text in conjunction with numeric data and can record and display previously made calls.
  • the telephone also has an answering capability which can screen incoming calls and can forward a particular call and display the associated name, if any, of the caller from its directory.
  • a novel speed dialing feature is also disclosed.
  • the telephone also has a novel line status monitoring circuit with a novel phone network interface circuit.
  • Figure 1 is a perspective view of a novel telephone device.
  • Figure 2 is a top view of the keyboard layout of the portion of the telephone shown in Figure 1.
  • Figure 3 is a schematic circuit block diagram of the telephone shown in Figure 1.
  • Figure 4 is a detailed block level schematic diagram of a portion of the circuit shown in Figure 3.
  • FIG. 5 is a detailed circuit level diagram of the microprocessor and its associated circuitry used in the telephone shown in Figure 3.
  • Figure 6 is a detailed schematic circuit diagram of the keyboard and display assemblies shown in Figure 3.
  • Figure 7 is a detailed schematic circuit diagram of the telephone network interface circuit portion of the telephone.
  • FIG. 8 is a detailed circuit level diagram of the interface circuit portion of the telephone which interfaces with various audio input/output signals.
  • Figure 9 is a graph showing dynamic time warping algorithm with a constant width band.
  • FIG. 10 is a graph showing adaptive linear pruning. Detailed Description of the Drawings
  • the telephone 10 comprises a hand set 12 which has a microphone and speaker.
  • the telephone 10 also has a numeric keypad 22 which receives the numeric input numbers (0-9) as well as the control signals of "*" and "#".
  • Such a keyboard is well known in the art.
  • the telephone 10 comprises a plurality of well known conventional control keys for redial 20, hold 18, flash 16, and speaker 14, to activate the speaker phone.
  • the telephone 10 also comprises a key labeled directory 24, a key labeled voice 26, and three reprogrammable option buttons 28 (A-C) .
  • the telephone 10 comprises a LCD display 30 which can display two line of alphanumeric characters each line with sixteen characters. Referring to Figures 3 and 4, there is shown in block level diagram the telephone 10.
  • the telephone 10 is connected to a telephone line comprising of a tip and ring 32.
  • the telephone line 32 is connected to a line protection circuit 34 which is then connected to a polarity guard circuit 36.
  • the signals (to and from) are separated by a hybrid circuit 40 (discussed in greater detail hereinafter) , from which the transmit and receive signals are supplied to the audio connect circuit 46 and the receive attenuator circuit 44, respectively.
  • a hybrid circuit 40 discussed in greater detail hereinafter
  • Other analog circuits which connect the handset receiver 50 and handset microphone 52 and the speaker phone microphone 54 are also shown in Figure 4 and are well known in the art.
  • the signals in the phone circuit shown in Figure 4 are supplied to and from a microcomputer 60 shown in Figure 3.
  • the microcomputer 60 is a 50943 microcomputer made by Mitsubishi.
  • the microcomputer 60 is based on the 6502 processor.
  • the microcomputer 60 contains internal memory in the form of RAM and ROM. In addition, it has timers.
  • the microcomputer 60 also provides bidirectional digital I/O ports.
  • the microcomputer 60 has a built in A/D converter with multiple multiplexed inputs.
  • the microcomputer 60 has a pulse width modulator capable of generating analog signals when the proper low
  • the timing of the microcomputer 60 is controlled by a crystal oscillator circuit 62.
  • the crystal oscillator circuit 62 consists of an 8 MHz quartz crystal and the support components. This is actually driven to oscillation by the microcomputer 60 which divides the resulting signal to get 20/OTJT and 0/OUT.
  • the signal 20/OUT is a 4 MHz digital clock signal that is used to drive the time control circuits 64.
  • the 0/OUT signal which is a 2 MHz digital clock signal, is used by the microcomputer 60 to set the basic processor cycle time. It is also supplied to a memory control circuit 66 which is used by the memory control circuit 66 to control access to external memory.
  • External memory in the form of 32K X 8 SRAM 68, 32K X 8 EPROM memory 70, and output latch 72 are all accessed by a 16 bit address bus from the microcomputer 60 with an 8 bit data bus. They are all under the control of the memory control circuit 66.
  • the memory circuit SRAM 68 is accessed whenever address 0000 through 7FFF are read from or written to.
  • EPROM memory 70 is accessed whenever address 8000 through FFFF are read from.
  • the digital output latch 72 is accessed whenever address 8000 through FFFF are written to. A system reset will initialize the digital output latch 72 causing all outputs to be set to a low logic level.
  • the output latch 72 is driven by the microcomputer 60 and can store the results from microcomputer 60.
  • the output latch 72 drives the following signals:
  • DTMF enable 74 This signal is supplied to the DTMF decoder 76. A logic level high on this line enables data output on the DTMF decoder circuit 76.
  • the LCD enable signal 78 is supplied to the LCD module 30.
  • a logic level high on the LCD enable 78 enables the LCD controller in the LCD module 30 to read and to write to the LCD display 30.
  • SYNTH 82 This signal is supplied to the synthesis control circuit 84. A logic level high will enable the output of the microcomputer 60 pulse with modulator to be sent to a low pass filter 86 for synthesis of analog audio signals.
  • RING signal 88 A logic level high on this line switches the output of the PWM driven low pass filter 86 to the ring drive path, permitting an audio signal to be supplied to the amplifier 90 of the speaker phone and to the speaker 92 of the speaker phone. This signal is used to synthesize the call ringer.
  • a low logic level on the RING line 88 routes the output of the low pass filter 86 to the speaker phone IC 48. From the speaker phone IC 48, the synthesized audio signal is sent through the audio connect circuit 46 to the tip and ring line to the telephone network to another phone and is used with the call answering feature.
  • LINE signal 94 This signal is supplied to the Hook switch 42 and controls the line status of the telephone 10. A logic level high takes the phone off hook. A logic level low makes the phone on-hook.
  • SPEAKER signal 96 This signal controls whether the phone is in normal, or speaker phone mode. A logic level high turns the speaker phone on. A logic low returns the operation to handset.
  • MUTE signal 98 A logic level high on this line causes attenuation of the receive signal coming from the telephone network of tip and ring 32. It also is used by the speaker phone IC 48 to attenuate the microphone amplifier 90. A logic level low allows normal signal levels.
  • a secondary function of the mute signal 98 is to control the selection of source signals for analog to digital conversion. When mute signal 98 is at a high logic level, signals coming from the telephone network 32 are sent to the recognition source selector 102 circuit which is then supplied through a low pass filter 104, through a sample and hold circuit 106 and to the analog to converter circuit within the microcomputer 60. When mute signal 98 is a logic low, signals from the microphone (handset 52 or speaker 54) are sent to the analog/digital converter in the microcomputer 60.
  • AUDIO signal 100 This signal controls the connection of the Audio connect circuit 46 to the tip and ring 32.
  • a logic level low allows normal operation.
  • a logic level high prevents audio signals from being transmitted or received from the telephone network 32. This is to implement the hold feature and also is used during the speech recognition process.
  • the microcomputer is also connected to the time control circuit 64.
  • the time control circuit has three functions: system reset, watchdog timer reset, and time reference interrupt. During power up, a reset pulse is generated. This is stretched to allow the microcomputer 60 to become stable. Manual reset is also stretched.
  • the 4 MHz signal, 20/OUT is divided down to a 61 Hz. (16.384 msec, period signal.) The resulting signal is used to drive the intl interrupt input on the microcomputer 60.
  • the watchdog signal 110 must be pulsed to a logic level low and then brought back high. This keeps the watchdog timer from causing an automatic reset of the microcomputer 60. If the watchdog circuit is left at a low logic level, the watchdog timer is disabled.
  • the microcomputer 60 also directly outputs or reads the following signals: l. Watchdog signal 110.
  • the watchdog signal 110 is supplied to the time control circuit 64.
  • a logic level high signifies normal operation. Pulsing the signal from high to low then back to high again is required once every 61 Hz interrupt to prevent a watchdog timer reset. Holding the watchdog signal 110 low disables the watchdog timer. The watchdog timer is used to insure that the microcomputer 60 is operations.
  • Battery signal 112. This is a bidirectional digital signal. This signal is normally used as an input to sense power supply status.
  • a logic level low that is read on this line by the microcomputer 60 indicates that power is being supplied to the telephone 10 by an AC transformer.
  • a logic level high read on this line indicates that batteries are powering the telephone 10.
  • Sense hook switch signal 116 This is a digital input signal to the microcomputer 60. This signal is used to detect the status of hook switch 42. A logic level high indicates that the telephone 10 is on hook. A logic level low indicates that the telephone 10 is off hook.
  • Serial in and serial out signals 118A and 118B These digital signals form an asynchronous serial communication port. This is used during the testing of the telephone 10. 5.
  • S/*H signal 120 This signal is supplied from the microcomputer 60 through the synthesis control circuit 84. This signal is used to drive the sample and hold switch circuit 106 which is used to supply input signal to the A/D converter portion of the microcomputer 60. A logic level high allows sampling of the signal from the sample and hold circuit 106. A logic level low prevents the receipt of signals from the sample hold circuit 106 into the microcomputer 60. When the signal is gated by a logic level high signal on SYNTH signal 82, it is also used to drive the low pass filter 86 that generates audio signals.
  • Slow bus (SB0-SB5) 122 This is a bidirectional bus for digital signals. It is a medium speed data and control bus for operating the keyboard 22 and the option buttons 28 (A-C) of the LCD module 30, and the DTMF decoder 76. SB2 through SB5 are data lines when communicating with the DTMF decoder 76 and the switches 28 (A-C) of the LCD module 30. SB0 and SB1 are control lines when communicating with the switches 28 (A-C) of the LCD module 30. SBO through SB5 are used as digital outputs to drive the keyboard 22.
  • ROWBUS ROW0-ROW3 124. These digital input signals are supplied from the keyboard 22 and are used to decode the keyboard key closures 22.
  • INT 126 This is a digital input signal received from the time control circuit 64. It is a 61 Hz interrupt signal that the microcomputer 60 uses to keep track of real time.
  • INT interrupt signal 128 This is a signal supplied from the DTMF decoder circuit 76 that indicates the presence of a valid DTMF tone.
  • Battery level signal 114 This is an analog input signal from the power supply 130. It is used to determine battery charge level.
  • Line status signal 132 This is an analog signal that is received by the microcomputer 60. It is used to detect incoming ring signals from the telephone network 32. It is generated by the line status monitor circuit 38. In addition, when the telephone 10 is on hold, this line is monitored to detect another telephone on the same line going off hook. If this occurs, the hold state will be ended.
  • Voice Signal 134 This analog input signal is supplied from the sample and hold circuit 106 to the microcomputer 60. Signals from the low pass filter 104 and the sample and hold signal 106 enter the A-D converter portion in the microcomputer 60. They are used for the speech recognition process and for software DTMF detection.
  • the hybrid circuit 40 which is used to interface the network telephone lines tip and ring 32 from the transmit and receive lines within the telephone, is shown as a single transistor 40.
  • the single transistor is of bipolar and shown as a PNP transistor Q5 MPSW63.
  • the PNP transistor has a collector 41, a base 39 and an emitter 37.
  • transmit audio coming from the CMOS switch (4053) 46, passes through the RC network C9, R10 to the base 39 of the transistor 40.
  • the audio signal into the base 39 modulates the collector current of transistor 40.
  • This collector current is the telephone loop current and is the transmit audio signal out to the telephone line 32.
  • the audio signal from the output of the CMOS switch 46 also passes through another RC circuit, C8 and R24, to the receive attenuator switch 44.
  • the transmit audio signal at the collector of transistor 40 is of equal amplitude but is one hundred eighty (180) degrees out of phase with the signal from 46.
  • a third RC network, Cll and Rll, passes this signal, summing it with the signal from the output of 46 causing a cancellation of the transmit audio into the receive attenuator 44. There is no cancellation of receive audio coming from the telephone line 32 which passes from the collector 41 of transistor 40 through this same RC network to the receive attenuator 44.
  • the integrated circuit designated Ul 4053 is a three pole double throw CMOS switch (44 & 46) . It is used to connect the receive audio path (C section) , and the transmit audio path (B section) , to the speaker phone IC.
  • the A section is a receive attenuator switch. It is used to mute the level of the DTMF signals and pulse clicks during dialing.
  • the line status monitor circuit 38 is a differential amplifier with a very high input impedance (greater than 10 megohms) .
  • the inputs are connected to the voltage out of the Polarity Guard 36.
  • the voltage is about 48 volts.
  • the op amp 38 (U3D) converts this voltage to a signal in the range of three volts and passes this to the line status input to the microcomputer 60.
  • the output of the Polarity Guard 36 is greater than 100 volts.
  • the output of the op amp 38 is saturated high (greater than 4 volts) .
  • the voltage at the output of 36 drops to a much lower value, in the region of 10 to 15 volts, which translates to a voltage less than 1 volt into the line status input.
  • These voltages are interpreted by software in the microcomputer 60 to determine line status.
  • the voltage change which takes place when another phone comes across the telephone line 32 is used by the software to terminate the hold function and drop the line when the second phone picks up the line.
  • the software to control the operation of the telephone 10 is stored in the ROM portion of the microcomputer 60 as well as the
  • EPROM memory 70 The software that is stored in the ROM section of the microcomputer 60 performs the functions of 1) CACF signal processing; 2) Low level hardware support routines; 3) LCD display text messages; and 4) Copyright and code protection code.
  • the software that is stored in the EPROM memory 70 performs the functions of speech recognition and user interface.
  • the RAM memory 68 is used as a scratch pad and for storage of voice templates during the operation of the telephone 10.
  • the user can use the keypad 22 in its normal prior art operation for dialing a particular user's numbers.
  • the numbers are displayed on the display 30.
  • the redial key 20, hold key 18., flash key 16, and the speaker key 14 function in the normal prior art manner.
  • a telephone number shall mean a plurality of digits.
  • the operation of the telephone 10 would proceed as follows.
  • the date and time would be displayed on the display screen 30.
  • the date and time can be changed by pressing the option button C 28 (C) twice and following the prompt on the display device 30 to change the date and time.
  • the telephone 10 Since the telephone 10 responds to a particular speech command, the telephone 10 must be first trained to store the speech pattern of the particular speech to which it will respond. In that connection. the user must first train the telephone 10. The training mode is entered by the user lifting the handset and activating the voice key 26. A message prompting the user to speak is displayed on the display 30. The user then speaks a word or a command. That speech, converted into an analog signal by the handset mike 52, is received by the telephone 10 through the recognition source selector 102, through the low pass filter 104, the sample and hold circuit 106 into the microcomputer 60. The microcomputer 60 performs a number of functions based upon the software that is set forth in Exhibit B.
  • the analog speech or command is supplied at the sample rate of 7200 HZ., and is digitized to yield X(t) . A difference between each sample is taken. Thus, the signal after the first difference would be
  • the resultant signal, S(t), from the first difference is to eliminate DC signal.
  • the difference operation places a 6 db octave preemphasis in the speech and thus acts like a high pass filter.
  • the sampled signals S(t) are supplied to a frame buffer comprising of 144 storage locations.
  • 144 samples form one frame. Therefore, the frame rate is at 20 msec.
  • a well known processing technique of clipped autocorrelation function is performed on each frame.
  • the clipped autocorrelation function performs the operation as follows:
  • Each of the coefficients ⁇ m) represents a value of a portion of the speech pattern in time.
  • a standard end point determination technique to determine the beginning and the ending of the speech is also applied.
  • the software set forth in Exhibit B then checks each frame by frame to compress the signal therein - also based upon the well known prior art technique.
  • the user is prompted to speak at least twice, or until two words are spoken which are consistent with each other. An average is taken of the two words that the user spoke. This average is based upon a standard, well known technique.
  • the normalized coefficients A N (m), associated with each frame that is calculated, based upon the clipped autocorrelation function, is then stored as the speech pattern of the inputted speech.
  • the user is then prompted to enter the telephone number associated with the speech inputted name.
  • the user then enters the telephone number associated with the inputted speech.
  • the user presses the option button 28(C) which is associated with the text display of "done".
  • the telephone 10 prompts the user to enter the alphabetical text name that corresponds to the speech name that was audibly inputted into the telephone 10.
  • the user simply presses the appropriate numeric key which contains the alphabetical letter.
  • the three choices are then displayed on the display 30.
  • Each choice is displayed juxtaposition to each one of the three option buttons 28 (A-C) .
  • the option buttons 28 (A-C) are then reprogrammed such that the activation of one of the keys would then input the particular associated displayed alphabetical letter on the display 30.
  • alphabetical text can be entered on the numeric keypad 22 in conjunction with the option buttons 28 (A-C) . For example, if the number "5" on the numeric keypad 22 is pressed and the control key 28A is pressed, the letter J would be entered into the telephone 10 and would also be displayed on the display 30.
  • the telephone 10 can store up to 50 speech names, each associated with a telephone number, and an alphabetical text name. Clearly, through the addition of greater memories, more names can be stored in the telephone 10.
  • the option buttons 28 (A- C) can be reprogrammed by the telephone 10 to function for other purposes.
  • the software to perform that function is contained in the listing set forth in Exhibit C.
  • the option buttons 28 (A-C) can be changed from changing the date and time function to change inputting alphabetical text function.
  • the telephone 10 can dial numbers in a conventional manner.
  • the telephone 10 can be placed in a mode whereby the keypad 22 is locked thereby preventing all outgoing calls.
  • each of the three option buttons 28 (A-C) would still be functional and they can be reprogrammed to be used for dialing the emergency numbers such as police, fire and ambulance.
  • the telephone 10 can be placed in a protected mode whereby the speech names and each name's associated telephone number and alphabetical text are protected from searches (further discussed) or deletions through reinputting.
  • the telephone 10 can also respond to speech command dialing.
  • the user picks up the handset 12 and simply speaks a name that the user wants to call - with the speech having been previously trained and stored in the telephone 10.
  • the speech is converted into an analog signal and is received again by the microcomputer 60 through the sample and hold circuit 106.
  • the microcomputer 60 once again performs the function of finding the first difference of the samples which are sampled at the rate of 7200 HZ. A frame of every 144 samples is also gathered together.
  • a clipped autocorrelation function of each frame is calculated and normalized.
  • the coefficients generated by the clipped autocorrelation function for the inputted speech is then generated.
  • the plurality of coefficients of the inputted speech pattern is then compared to the plurality of coefficients associated with a stored speech pattern.
  • the comparison is based upon a modified dynamic time warping (DTW) algorithm.
  • DTW dynamic time warping
  • Figure 9 shows a typical band Dynamic Programming graph (see "Dynamic
  • the band region is defined as:
  • the processed spoken word i.e., the coefficients of the clipped autocorrelation of the spoken word is compared to each one of the stored words using the DTW algorithm and the word whose coefficients produce the smallest DTW result is the closest word to which the input word correspond.
  • the closest match i.e., the result of the DTW operation is still greater than some threshold level, then no match is found.
  • the matching of the speech pattern of the spoken word to the speech pattern of the stored words must be made through the list of the speech patterns of all the stored words before the correct match is found.
  • an adaptive linear pruning method is employed in the searching or matching process.
  • the DTW operation is first operated on the first word.
  • the DTW operation computes not only the result of the operation but also every subpart summation that corresponds to the particular point in time (see Figure 10) .
  • (Thus, C*, - C N are computed) .
  • the line 200 represents the best result of a DTW operation on a word, then not only is the DTW value for the total (C H ) computed, but also the linear progressive coefficients (C-
  • a second word comparison would be made at each point in time between the coefficient as a result of the DTW operation of the spoken word to the second word and to the corresponding coefficients for the best word.
  • line 210 represents.the operation of the DTW algorithm on the spoken word compared to the second word, then the operation is terminated without waiting for the completion of the DTW operation on the entire coefficients that correspond to the second word.
  • the predictive nature of the progression of the summation of the differences in line 210 will get worse and that the operation need not be permitted to its completion. This, of course, saves computational time and speeds the searching result.
  • an offset "O" can be added to the best score.
  • the DTW operation on other words in subsequent operations, as each coefficient in time is presented, must exceed the best score plus the offset to ensure that the prediction of the calculation of DTW through completion would in all likelihood exceed the DTW of the best score.
  • abs_mthresh This is absolute match threshold that a match must be under to be considered as a valid match.
  • rel_mthresh This is the relative match threshold over which the best match must be greater than the second best match to not be considered “questionable.”
  • a match is considered "pruned" when D n (x) > PL(x) , where D n (x) is the normalized accumulating distance function of the DTW.
  • the prune threshold is set to a maximum absolute cutoff threshold initially. If the results of the DTW operation on the first word is such that it is below the initial maximum threshold, then all of the coefficients of the first word would have been operated upon by the DTW operation. Thereafter, the probability decreases as to whether or not the DTW operation is allowed to operate on all of the coefficients of the subsequent words.
  • One possible method as used by the telephone 10 is to present words which correspond to names and therefore a telephone number which has been dialed most frequently. This raises the probability that the spoken word would most likely match the stored word which corresponds to the most frequently dialed telephone number.
  • a pre-sort feature of the stored words is accomplished by the telephone 10 before the stored words are presented for the DTW adaptive pruning analysis.
  • a general purpose microcomputer 60 can be used in the telephone 10. This saves cost in the telephone 10. Because it is always possible that the telephone 10 is unable to match the speech pattern of the spoken word with the speech pattern of any of the stored words, the stored words that are most probable of being matched, e.g., the stored words that have the lowest values of DTW algorithm operation but still exceed a threshold amount, are presented for display on the display 30. A match j that falls into the "questionable" region must satisfy the following criterion:
  • the telephone 10 presents its best questionable match for the user to confirm whether it is indeed the word that was spoken.
  • the user can then press option button A 28 (A) ("yes") if it is correct, or option button B 28 (B) ("no") to request the telephone 10 to display the next best questionable match. This process continues until either "yes” (28 (A) ) is pressed or there is only one remaining questionable match left, wherein it is dialed. Thus, on questionable matches, user input is sought.
  • Another way for a user to use the telephone 10 is to press the directory key 24.
  • the user is then prompted to enter one of the numbers from the numeric keypad 22.
  • the number that the user selects has three alphabetical letters associated therewith. All the names in that three letter group are presented in alphabetical order. To scroll forward, the user presses that same numeric key repeatedly. When a particular desired name and number are presented on the display 30, the user then activates the redial key 20 to dial that number.
  • the software to accomplish this is set forth in Exhibit C.
  • the telephone 10 can also be used to retrieve a telephone number - without dialing. There are two methods to accomplish the foregoing. In one method, the user simply presses the directory key 24 and then presses the voice button 26. The user then speaks the name. The telephone 10 will process the speech signal as previously discussed and display the chosen name and the telephone number associated with that name.
  • the telephone 10 can also be searched to retrieve a telephone number manually.
  • the user presses the directory key 24 and then a number on the keypad 22.
  • the user does not have to press an option button 28 (A-C) to narrow a three (3) letter group down to one.
  • the user presses that same key repeatedly to move forward.
  • the option buttons 28 (A-C) are not used at all except to display the one-button speed dial names/numbers.
  • the software to accomplish this is set forth in Exhibit C.
  • the telephone apparatus 10 can also log the last one hundred calls made.
  • the telephone 10 also stores the time, the date, and the phone number called and the length of each call.
  • the user can review this log at any time to audit phone bills, scan for frequently called numbers or redial those numbers, or any other purpose.
  • option button B 28 (B) The software to accomplish this is set forth in Exhibit C.
  • the telephone 10 also has an answering facility. It can be activated by pressing option button A 28 (A) . When it is activated, in the unattended mode, the telephone 10 will answer all calls with a prerecorded message. The apparatus 10 then prompts the caller to use the caller's telephone to dial in the telephone number of the caller. The caller, however, must be at a telephone apparatus that generates a DTMF signal. The DTMF signals are received by the telephone 10 and are converted into signals that represent the telephone number of the caller. The telephone 10 then records that number and the time of the call.
  • the telephone 10 can use the telephone number received from the caller and search through its directory to find the corresponding name. When the user returns, the telephone 10 can display the number of messages it has recorded. In addition, the time and the phone numbers are displayed. Finally, if there is a name associated with the phone numbers, i.e., the phone numbers are in the directory of the telephone apparatus 10, the name will be displayed as well.
  • Another aspect of phone answering is that the user can program the telephone 10 such that once a call has been answered and the caller has left the message comprising the DTMF signals indicating the telephone number of the caller, the telephone apparatus can then automatically dial a preset number (which is the telephone number of a paging service) and regenerate the DTMF signals left by the caller.
  • the telephone apparatus 10 can relay the numbers of the caller after a caller has left the message of where the caller can be reached.
  • the telephone 10 can be placed in call screening mode.
  • this is activated by pressing option button A 28 (A)
  • the telephone apparatus 10 does not ring when a incoming call is received.
  • the calling party knows the preassigned code, the calling party can dial in the code. This would override the call screening capability and the ringer would then be turned on. Without the code, the calling party then receives the phone answering message and the telephone 10 records the calling telephone number of the caller.

Landscapes

  • Engineering & Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Signal Processing (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Telephonic Communication Services (AREA)
  • Telephone Function (AREA)

Abstract

Appareil de traitement de la parole. Dans un mode de réalisation, l'appareil est utilisé pour activer un téléphone. Le téléphone activé par la parole enregistre des schémas vocaux selon un algorithme écrêté modifié de fonction d'autocorrélation. La comparaison entre le schéma vocal des mots prononcés à haute voix et le schéma vocal des mots enregistrés afin d'obtenir une correspondance entre des schémas vocaux est effectuée selon un algorithme modifié de distorsion temporelle dynamique, une fenêtre de largeur constante étant maintenue. En outre, on utilise un procédé adaptatif d'élagage afin d'accélérer l'opération des algorithmes de distorsion temporelle dynamique. Une pluralité de mots prononcés à haute voix, le numéro du téléphone et le mot alphanumérique associé à chaque mot prononcé à haute voix sont enregistrés dans le téléphone. Le téléphone compose automatiquement un numéro de téléphone en réponse à l'entrée d'un mot prononcé à haute voix, après avoir retrouvé le mot prononcé enregistré correspondant. En outre, le numéro de téléphone et le texte alphanumérique correspondant au mot prononcé sont affichés.
EP19900902869 1989-01-05 1990-01-04 A speech processing apparatus and method therefor Withdrawn EP0453511A4 (en)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US29387689A 1989-01-05 1989-01-05
US07/294,168 US5007081A (en) 1989-01-05 1989-01-05 Speech activated telephone
US293876 1989-01-05
US294168 1994-08-22

Publications (2)

Publication Number Publication Date
EP0453511A1 true EP0453511A1 (fr) 1991-10-30
EP0453511A4 EP0453511A4 (en) 1993-05-26

Family

ID=26968205

Family Applications (1)

Application Number Title Priority Date Filing Date
EP19900902869 Withdrawn EP0453511A4 (en) 1989-01-05 1990-01-04 A speech processing apparatus and method therefor

Country Status (4)

Country Link
EP (1) EP0453511A4 (fr)
JP (1) JPH04504178A (fr)
KR (1) KR910700582A (fr)
WO (1) WO1990008439A2 (fr)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
NL1000284C2 (en) * 1995-05-02 1996-11-05 Huigh Cornelis Van Der Mandele Voice operated number selection unit for telephone
GB2307137B (en) * 1995-11-04 2000-03-22 Motorola Ltd A communications addressing network and terminal therefor
EP0778689A1 (fr) * 1995-12-06 1997-06-11 WILHELM, Siegfried E Terminal de télécommunication
US5960395A (en) 1996-02-09 1999-09-28 Canon Kabushiki Kaisha Pattern matching method, apparatus and computer readable memory medium for speech recognition using dynamic programming
US5892813A (en) * 1996-09-30 1999-04-06 Matsushita Electric Industrial Co., Ltd. Multimodal voice dialing digital key telephone with dialog manager
US6233557B1 (en) 1999-02-23 2001-05-15 Motorola, Inc. Method of selectively assigning a penalty to a probability associated with a voice recognition system
GB0624085D0 (en) * 2006-12-01 2007-01-10 Oxford Biosignals Ltd Biomedical signal analysis method

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1984003983A1 (fr) * 1983-03-28 1984-10-11 Exxon Research Engineering Co Procedes et dispositifs de reconnaissance de la parole

Family Cites Families (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3746793A (en) * 1972-08-09 1973-07-17 Phonics Corp Telephone communication system for the hearing impaired
US4074069A (en) * 1975-06-18 1978-02-14 Nippon Telegraph & Telephone Public Corporation Method and apparatus for judging voiced and unvoiced conditions of speech signal
US4400589A (en) * 1979-05-21 1983-08-23 United Networks, Inc. Subscriber station network
US4425627A (en) * 1981-02-23 1984-01-10 Sperry Corporation Intelligent prompting terminal apparatus
CH644246B (fr) * 1981-05-15 1900-01-01 Asulab Sa Dispositif d'introduction de mots a commande par la parole.
JPS5844500A (ja) * 1981-09-11 1983-03-15 シャープ株式会社 音声認識方式
GB2118099B (en) * 1982-02-23 1986-12-03 Confon Ag Telephone card index with automatic dialing
US4536886A (en) * 1982-05-03 1985-08-20 Texas Instruments Incorporated LPC pole encoding using reduced spectral shaping polynomial
US4608457A (en) * 1984-04-11 1986-08-26 Fowler Stephen L Telecommunications device for the hearing impared
JPS60236554A (ja) * 1984-05-10 1985-11-25 Sharp Corp 電話機
US4644107A (en) * 1984-10-26 1987-02-17 Ttc Voice-controlled telephone using visual display
US4783803A (en) * 1985-11-12 1988-11-08 Dragon Systems, Inc. Speech recognition apparatus and method
DE3607728A1 (de) * 1986-01-25 1987-07-30 Telefonbau & Normalzeit Gmbh Endgeraet einer fernmeldevermittlungsanlage
US4741031A (en) * 1986-06-27 1988-04-26 Gai-Tronics Intrinsically safe telephone
US4866778A (en) * 1986-08-11 1989-09-12 Dragon Systems, Inc. Interactive speech recognition apparatus
JP2584249B2 (ja) * 1986-10-31 1997-02-26 三洋電機株式会社 音声認識電話機
US4827500A (en) * 1987-01-30 1989-05-02 American Telephone And Telegraph Company, At&T Bell Laboratories Automatic speech recognition to select among call destinations
US4751728A (en) * 1987-03-27 1988-06-14 Treat John M Telephone call monitoring, metering and selection device

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1984003983A1 (fr) * 1983-03-28 1984-10-11 Exxon Research Engineering Co Procedes et dispositifs de reconnaissance de la parole

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
ICASSP'81 (IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING, Atlanta, Georgia, 31st March - 1st April 1981), vol. 1, pages 188-192, IEEE, New York, US; H. NEY: "Telephone-line speaker recognition using clipped autocorrelation analysis" *
IEEE TRANSACTIONS ON AUDIO AND ELECTROACOUSTICS, vol. AU-16, no. 2, June 1968, pages 235-239, New York, US; R.F. PURTON: "Speech recognition using autocorrelation analysis" *
See also references of WO9008439A3 *

Also Published As

Publication number Publication date
JPH04504178A (ja) 1992-07-23
KR910700582A (ko) 1991-03-15
WO1990008439A3 (fr) 1990-09-07
EP0453511A4 (en) 1993-05-26
WO1990008439A2 (fr) 1990-07-26

Similar Documents

Publication Publication Date Title
US5007081A (en) Speech activated telephone
EP0311414B1 (fr) Dispositif de numérotation à commande vocale avec mémoires pour la numérotation complète par tous les usagers et numérotation abrégée pour les personnes autorisées
EP0307193B1 (fr) Appareil téléphonique
KR100616787B1 (ko) 전화 핸드셋용 음성 인식 사용자 인터페이스
US5960393A (en) User selectable multiple threshold criteria for voice recognition
US6940951B2 (en) Telephone application programming interface-based, speech enabled automatic telephone dialer using names
US5752230A (en) Method and apparatus for identifying names with a speech recognition program
GB2317781A (en) Multimodal voice dialling digital key telephone with dialog manager
US6563911B2 (en) Speech enabled, automatic telephone dialer using names, including seamless interface with computer-based address book programs
EP0893901B1 (fr) Méthode de contrôle d'un service de télécommunication et d'un terminal
CA2221913A1 (fr) Correction de numeros de compte alphanumeriques au moyen d'une base de donnees statistiques pour la reconnaissance de signaux vocaux et de tonalites
US6671354B2 (en) Speech enabled, automatic telephone dialer using names, including seamless interface with computer-based address book programs, for telephones without private branch exchanges
USH1646H (en) Speech recognition adapter for telephone system
EP0453511A1 (fr) Procede et appareil de traitement de la parole
US6223161B1 (en) Method for setting terminal specific parameters of a communication terminal
CA2058644C (fr) Poste telephonique actionne par la voix
KR930005223B1 (ko) 음성 인식장치 및 이것을 이용한 전화기
US20040037399A1 (en) System and method for transferring phone numbers during a voice call
KR970055729A (ko) 이동단말기에서의 음성인식에 의한 전화번호 송신방법 및 장치
JPH06311220A (ja) 画像認識ダイヤル装置
EP0642118B1 (fr) Système automatique pour l'obtention guidée de signaux de parole sur une ligne téléphonique
KR950009425B1 (ko) 음성인식 전화기의 다이얼링 제어 방법
US20040181415A1 (en) Speech activated telephone device for connection to existing landline phone
CA2256781A1 (fr) Methode et appareil pour composer automatiquement un numero de telephone voulu, au moyen de commandes vocales
JPH06303320A (ja) Ｐｂｘ装置

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 19910627

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AT BE CH DE DK ES FR GB IT LI LU NL SE

A4 Supplementary search report drawn up and despatched

Effective date: 19930405

AK Designated contracting states

Kind code of ref document: A4

Designated state(s): AT BE CH DE DK ES FR GB IT LI LU NL SE

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 19930803