KR940010226B1 - Speech recognition and responding system for carphone - Google Patents

Abstract

The sound recognition and response system for a car phone effectively removes noise in an environment where there is low frequency noise to maintain a regular recognition rate regardless of microphone kinds, its position and a speaker's voice level. The system includes a microphone for inputting a voice signal, a first amplifier, an impedance matching part for matching the impedance between the microphone and the first amplifier, an automatic level controller, a low frequency blanking filter for removing noise of the low frequency, an A/D converter, a controller, a program memory, a data memory, a D/A converter, a second amplifier, a carphone, and a UART for interfacing the controller and the carphone.

Description

Voice recognition and response system for car phone

1 is a block diagram of a conventional car speech recognition and response system.

2 is an overall block diagram of a voice recognition and response system for a car phone according to the present invention.

3 is a detailed circuit diagram of an analog signal processing apparatus according to FIG.

4 is a detailed block diagram of a digital signal processing apparatus according to FIG.

5 is a detailed circuit diagram of the interrupt generator employed in FIG.

6 is a detailed circuit diagram of a time delay circuit employed in FIG.

* Explanation of symbols for main parts of the drawings

1, 10: microphone 2, 130: microphone amplifier

3: voice recognition device 4: control unit

5, 220: program memory section 6, 230: data memory section

7, 260: UART 8: Voice recording and synthesis device

9, 300: car phone 100: analog signal processing device

120: impedance matching unit 140: automatic gain control unit

150: low frequency cancellation filter unit 160: anti-aliasing filter unit

170: A / D converter 200: digital signal processing device

210: signal processor 240: D / A converter

250: speaker amplification unit 261: bidirectional buffer

262: decoder 271: latch circuit

272: latch control circuit 273: interrupt generator

274: time delay circuit 275: clock generator

276: backup battery circuit

The present invention relates to a voice recognition and response system for a car phone, and in particular, a car voice that can maintain a constant recognition rate regardless of the microphone type, location, and speaker's voice level by effectively removing the noise in the presence of low frequency noise. A recognition and response system.

Conventionally, as shown in FIG. 1, the audio signal input from the microphone 1 is amplified by the microphone amplifying unit 2, and then the voice is recognized by the voice recognition device 3 under the control of the control unit 4. FIG. Make it aware. Further, under the control of the control unit 4, the audio signal processed by the separate dedicated voice recording and synthesizing device 9 is output to the car phone 9. However, the conventional voice recognition system as described above has changed remarkably according to the position, type, and voice level of the microphone, and the voice level (75dB to 82dB) in a limited space such as in a vehicle, that is, low frequency noise. Compared to the noise level (70dB ~ 72dB) is not much larger, the signal-to-noise (S / N) ratio is low, there is a problem that the recognition rate is lowered.

In addition, there is a problem of bulkiness by using a separate recording device and a synthesizing device for the response function, and when there is no response function, the driver performs dialing while watching the LCD (Liquid Crystal Device) screen while driving. There were other problems, such as being unable to guarantee and causing accidents.

SUMMARY OF THE INVENTION The present invention has been made to overcome the above problems, and provides a voice recognition and response system for a car that guarantees the driver's safety when the driver wants to dial while driving and improves the voice recognition rate in the car. That is the purpose.

In order to achieve the above object, a car phone, a microphone for inputting a voice signal, a microphone amplifier for amplifying the voice signal output from the microphone, an impedance matching unit for matching the impedance between the microphone and the microphone amplifier; And an automatic level controller for keeping the voice level constant regardless of the speaker's voice level and microphone position or type, and for removing low frequency noise from the signal output from the auto level controller. A low frequency cancellation filter unit, an anti-aliasing filter unit for eliminating the aging generated when the audio signal is restored, and an A / D converter for converting the analog audio signal output from the anti-aliasing filter unit into a digital signal. And digitally outputting the digital signal output from the A / D converter. A signal processing processor for controlling the system for point detection, amplitude calculation, calculation of linear predictive coding coefficients, pitch detection and pattern matching and determination, a data memory unit for storing the processed signal from the signal processing processor, and controlling the system A program memory unit having a built-in program, a D / A converter for restoring the voice data processed by the controller to an analog signal, and amplifying the signal output from the D / A converter for output to a speaker And a speaker amplifier and a UART for interfacing the car phone with the signal processor.

Hereinafter, a voice recognition and response system for a car phone according to the present invention will be described with reference to the accompanying drawings.

2 is an overall block diagram of a voice recognition and response system for a car phone according to the present invention.

Referring to FIG. 2, the analog signal processing apparatus 100 includes a microphone 110, an impedance matching unit 120, a microphone amplifier 130, an automatic level controller 140, a low frequency cancellation filter 150, and an anti- It includes an aging filter unit 160, A / D conversion unit 170, the digital signal processing apparatus 200 is a signal processing processor 210, a program memory unit 220, a data memory unit 230, A D / A converter 240, a speaker amplifier 250, and a UART 260 are included, and the car phone 300 includes a handset 301, a wireless receiver 302, and a speaker 303.

3 is a detailed circuit diagram of the analog signal processing apparatus 100 shown in FIG.

Referring to FIG. 3, the microphone 110 is connected to an electric diaphragm ED at an input terminal of the impedance converter CON. The voice signal input from the microphone 110 is a very fine signal of several μV to several tens of mV. The impedance matching unit 120 looks at the impedance seen from the output terminal of the impedance converter CON and the input terminal of the microphone amplifier 130 to minimize the distortion of the voice signal regardless of the type of the microphone 110. The impedance may be varied by adjusting the voltage by the resistors R1, R2, and R3 to match the impedance. In the microphone amplifying unit 130, a high pass filter 131 composed of a capacitor C4 and a resistor R4, a low pass filter 132 composed of a resistor R5 and a capacitor C5, and resistors R6, The audio signal passed through the non-inverting amplifier 133 composed of the operational amplifier OP1 whose gain ratio is adjusted by R7) has a passband of 160 Hz to 72 kHz and has an amplification degree of about 3 times. In the automatic level controller 140, the diode D1 and the capacitor C1 are half-wave rectifiers for reducing the dynamic range of the signal.

The rectifier 142 is a full-wave rectifier to adjust the gain control current I to the gain cell ΔG according to the signal level passing through the rectifier 142. The gain cell ΔG supplies a current fed back to the summing node A of the operational amplifier OP2 according to the current amount I. Resistor R14 and condenser C9 form an alternating current feedback loop, and resistors R12 and R13 and condenser C8 form a direct current feedback loop. In addition, the gain is changed by the resistors R10 and R11, and the resistors R8 and R9 are bias resistors for adjusting the DC level of the output signal. The 1.8V power supply connected to the non-inverting terminal of the operational amplifier OP2 is a bias voltage to prevent signal distortion.

That is, the automatic level control unit 140 is for limiting the AC peak value of the voice signal to between -5V and + 5V, so that if the peak value of the signal is within 10V, a large amount of current is supplied to the gain cell ΔG so that the level of the signal is limited. If it is 10V or higher, supply a small amount of current to the gain cell ΔV to lower the level to maintain a constant level.

The high pass filter 151 and the band pass filter 152 of the low frequency cancellation filter unit 150 are hybrid ICs that remove noise of 300 Hz or less from the output signal of the automatic level control unit 140.

The anti-aliasing filter unit 160 includes a low pass filter 161 and a clock generator 162 for generating a 460 kHz clock. In the low pass filter 161, since 1/100 of the main clock is a cut-off frequency, 4.6 kHz is a cutoff frequency of the low pass filter 161.

That is, since the A / D converter 170 samples the voice signal at 10 kHz, the maximum frequency of the input signal must be limited to within 5 kHz by the Nyquist rate in order to restore the signal without distortion. A low pass filter 161 having a frequency is used.

In addition, the anti-aliasing filter unit 160 adjusts the DC offset value of the A / D converter 170.

The A / D converter 170 converts the signal output from the anti-aliasing filter unit 160 into a 12-bit digital signal, passes through the latch circuit 271, and then outputs the signal to the signal processing processor 210. . The latch circuit 271 receives a control signal of the latch control circuit 272 decoded by the decoder 261.

Meanwhile, the signal processing processor 210 needs to detect and detect the start and end points of the voice signal for voice recognition and learning. At this time, as shown in FIGS. 5A and 5B, the interrupt generator 273 uses a sampling frequency of 10 kHz in the A / D converter 170, that is, every 100 μsec sampling cycle having a high period of 90 μsec and a low period of 10 μsec. The moment that the audio signal is sampled, the A / D converter 170 A control signal is input from the terminal to the interrupt generator 273. remind Since the signal output from the terminal has a low duration of 10 μsec, it is necessary to reduce the low duration to within the processing time of the interrupt subroutine. The signal processing processor 210 can operate when the low period of the interrupt signal becomes 200 nsec or more. In the system according to the present invention, the low period of the interrupt signal is set to about 500 nsec.

On the other hand, the interrupt generator 273 is composed of two D flip-flops, the clock signal and the interrupt recognition signal of 10MHz from the signal processing processor 210 is input to the or gate (G1), the or gate (G1) The output signal Q of the second D flip-flop outputs an interrupt signal having a low period of 500 nsec. That is, if the interrupt service routine time is longer than the low period of the interrupt signal on the program of the signal processing processor 210, there is no problem. However, if the interrupt service routine time is shorter than the low period of the interrupt signal, the interrupt may take several times. The interrupt generator 273 outputs an interrupt signal to the signal processor 210.

The digital signal processing processor 210 processes the digital signal output from the A / D converter 170 by a program embedded in the program memory unit 220 to process four 32K × 8-bit SRAMs ( The memory selection of the data memory unit 230, which is a static random access memory, is performed by the memory selection circuit 277 under the control of the signal processing processor 210.

Meanwhile, since the command cycle of the signal processing processor 210 is 100 nsec in the wait state generator 274, when the access time of the external device is slower than the command processing time of the signal processing processor 210, the signal processing processor 210 is performed. ) To access the external device.

As shown in FIG. 6, the present invention is composed of two JK flip-flops because a maximum of two waits are required. Devices requiring two waits are the A / D converter 170, the D / A converter 240, and the UART 260, and the chip select terminals of the devices. Is connected to the input terminal of the NAND gate G3, and the output of the NAND gate G3 is input to the input terminal J1 of the first JK f / f. Devices requiring one wait are erasable programmable read only memory (EPROM) of the program memory unit 220 and SRAM of the data memory unit 230, and chip select terminals of the devices. Is connected to the input terminal of the NAND gate G4, and the output of the NAND gate G4 is connected to the input terminal J2 of the second JK f / f.

A 10 MHz clock signal output from the signal processing processor 210 is input to the first and second clear stages CLR1 and CLR2, and a reset signal output from the signal processing processor 210. Is input to the first and second clear stages CLR1 and CLR2, is input to the NAND gate G4 of the inverted output stage Q1 of the first JK f / f, and the second JK f / f inverted output stage Q2. The output signal of is input to the ready terminal (READY) of the signal processing processor 210.

On the other hand, prior to performing the speech recognition, learning is required through the start detector 211 and the LPC coefficient calculator 213 of the signal processing processor 210.

At this time, the learning mode of the handset 301 of the car phone 300 may be selected to learn about 15 names and names, 5 commands, 10 numbers, and the like. After learning, the recognition mode is selected, and the recognized vocabulary is displayed on the LCD screen of the handset 301 of the car phone 300 by the UART 260 or the analog signal from the D / A converter 240. After the restoration is amplified in the speaker amplification unit 250 is output to the speaker 303 of the car phone 300. Here, the speaker amplifier 250 includes capacitors C12, C13, C14, and C15 and resistors R15 and R16 for matching the output impedance of the D / A converter 210 and the input impedance of the speaker 303. )

Therefore, by outputting an analog voice signal through the speaker 303, the driver can check the recognized vocabulary through the ear to promote the safety of the driver, and the wrongly recognized vocabulary can be canceled and the recognition can be performed again. The dial success rate can be increased.

As described above, the voice recognition and response system for a car phone according to the present invention has the effect of maintaining the recognition rate regardless of the microphone type, location, and voice level of the speaker even in a low frequency noise of 300 Hz or less. In addition, the user can check the recognized vocabulary through the LCD screen and the speaker by adding a response function, thereby improving driver safety.

The invention is also applicable to any device that desires a man to machine interface via voice.

Claims (5)

For matching the impedance between the microphone 110 for inputting a voice signal, the first amplifier 130 for amplifying the voice signal output from the microphone, and the microphone 110 and the first amplifier 130 An automatic level control unit 140 for maintaining a predetermined sound level regardless of the speaker's voice level and the microphone's position or type, with the impedance matching unit 120 and the signal output from the first amplifier 130; A low frequency cancellation filter unit 150 for removing low frequency noise from the signal output from the automatic level control unit 140, an anti-aliasing filter unit 160 for removing an ergonomic generated when the voice signal is restored, and the entity An A / D converter 170 for converting the signal output from the aliasing filter into a digital signal, a controller 210 for processing the digital signal output from the A / D converter 170,The program memory unit 220 embedded in the program for stirring the system, and the data memory unit 230 for storing the data output by processing the digital signal of the A / D conversion unit 170 in the control unit 210 And a D / A converter 240 for restoring the voice data processed by the controller 210 to an analog signal, and for amplifying a signal output from the D / A converter 240 to output to a speaker. Interfacing the second amplifier 250, the car phone 300 to check the dialing and the signal output from the second amplifier 250 with the eyes or ears, and the controller 210 and the car phone 300 Voice recognition and response system for a car, characterized in that it comprises a UART (260) for. According to claim 1, wherein the automatic level control unit 140 is a half-wave rectifier 141 for reducing the dynamic range of the audio signal output from the first amplifier 130 and the half-wave rectifier 141 And a detector 142 for detecting the level of the signal, and a level controller 143 for maintaining a constant level of the signal by converting the amount of current according to the signal detected by the detector 142. Recognition and response system. The voice recognition and response system for a car phone according to claim 1, wherein the low frequency cancellation filter unit (150) comprises a high pass filter and a low pass filter. The voice recognition and response system for a car phone according to claim 1, wherein the controller (210) comprises a time delay circuit composed of two JK flip flops. The voice recognition and response system for a car phone according to claim 1, wherein the control unit (210) includes an interrupt generator configured to generate two D flip-flops to generate an interrupt signal.