JPS6226480B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an audio control device for a television receiver, etc., and in particular, when issuing another command to an operating receiver, etc. to perform audio control, it is possible to In addition to increasing the recognition tolerance value, muting is applied to the audio circuit of the receiver using the identification (recognition) output, and the sound (acoustic) emitted from the receiver itself.
The purpose is to prevent malfunctions caused by output.

Commander (an operator, hereinafter referred to as a person who remotely controls a controlled device (for example, a television receiver, etc.) using real voice as source information). ] By standardizing and quantizing the standard voice command (or instruction) content (voice signal), it is stored in advance as a standard (digital signal) pattern, and the voice instruction content (voice signal) to be issued later is stored in advance as a standard (digital signal) pattern. After standardization, it is quantized, and if necessary, the time axis is adjusted with respect to the standard audio signal, digitized, and then temporarily stored in a memory such as RAM. When matched with a value, turns on.
It has been proposed to perform off control.

Examples of such voice recognition means include, as shown in FIG. 1, an input unit 1 including an acoustic-electrical signal converter (for example, a microphone) that converts input voice into an electric signal, and a feature extraction unit that extracts the characteristics of the voice signal. Department
2. Standard pattern storage unit 3 that stores standard patterns of voice features registered in advance; Recognition processing unit 4 that compares the characteristic patterns extracted from the input voice with the standard pattern and identifies the input voice; Based on the recognition results; For example, the main component is an output control section 5 that controls the power, channel, volume, etc. of a television receiver, and an input signal amplitude normalization circuit 6 and a time axis adjustment section 7 to improve the recognition rate. A registration control section 8 is added for registering standard patterns of audio features.

Many methods can be considered as parameters for extracting voice features, such as frequency spectrum distribution, correlation function, number of zero crossings, formal frequency, or linear prediction coefficient, but among these methods, the frequency spectrum of voice is separated using multiple frequency filters. The so-called filter bank method, which extracts a pattern and examines its correlation with a standard pattern, is often used as a method that can obtain a high recognition rate with a relatively simple configuration.

Controls within the control device using such voice include turning the power on and off, specifying changes to channels and station numbers,
For example, the power source can be controlled by saying things like ``dengen iri'' (kiri), but since the volume is originally an analog continuous variable control, it was difficult to control it by voice.
Regarding this method of controlling the analog control amount by voice command, the applicant and others have already filed a patent application filed in 1973-
The application has been filed as No. 59235, and the gist of the application is to numerically indicate such an analog quantity by voice, and the content is discriminated by a voice recognition device, for example, 10
The object of the present invention is to control the level of the analog quantity by extracting it as a binary value, and changing the combination of a plurality of variable attenuators or variable gain amplifiers, and to display the level.

By the way, when the television receiver is already operating, such as when adjusting the volume or changing channels while receiving a picture, the "1" and "2" in the sound emitted from the receiver itself …
Malfunctions may occur due to numbers such as .

In particular, we have increased the level of recognition requirements for American and Chinese words.
Such considerations are essential when attempting to reduce the variation in recognition due to the utterance conditions of the dispatcher by lowering the specific recognition rate and increasing the tolerance of speech recognition.

In view of this, the present invention proposes a method of mutating the audio circuit of a receiver, especially the circuit that drives the sound emitting means such as a speaker, when a voice command is given. The details of the present invention will be explained below with reference to FIG. 2 showing one embodiment.

This embodiment will be explained by taking as an example a voice control device for a television receiver incorporating a voice recognition device that employs a filter bank method (frequency spectrum method) as a feature parameter for voice recognition. Any controlled device may be used as long as it is equipped with a sound emitting means.

The input section 1 , which is usually attached to the front of the device, has two microphones 10, one directional and one omnidirectional.
and 11 as shown in the figure, and an amplifier 12. That is, the omnidirectional microphone 11 is connected in opposite phase to the directional microphone 10, and therefore, audio signals from outside the directional characteristic range, that is, signals other than control command voices, are canceled out, and signals within the directional characteristic range are canceled out. The signal-to-noise ratio of the control command voice is thereby increased. When the present invention receives a voice command,
In order to prevent malfunctions due to similar command words in the audio coming from the speakers of TV receivers, etc., important words (most important words) ``Dengen'' or ``Power'', ``Channel'', ``Onlyo'' or ``Volume'' are used. When a command such as the following (note: large tolerance values are set for these terms) is identified, a so-called initial mutating circuit, which is separately disclosed as a main circuit diagram in Figure 2, is created. Set it up,
Temporarily cut off the speaker output or significantly attenuate it. The feature extraction unit 2 , which also has an amplitude normalization function, includes a plurality of filters 13-1, 13-2...
13-N, a level detection circuit 14 that detects the total amplitude of the input signal, an A-D (analog-digital) converter 15 that converts the output of each filter into a digital signal, and An amplitude normalization circuit 16 configured with an analog divider or the like normalizes the filter output amplitude by taking the ratio of each filter output and the output of the level detection circuit 14; The multiplexer 17 is inserted between the conversion circuit and the filter group and switches the connection of each output of the filter. With such a configuration, each filter component of the audio signal input from the input section 1 is sequentially sampled at appropriate time intervals (often around 10 milliseconds), and further quantized by quantizing each sampling value. After being converted into digital code, a microcomputer or central processing unit (CPU) 1
The data is stored in a storage memory 19 (usually RAM: random access memory) through an I/O port (not shown) 8.

In the process of A-D conversion, when each sampled quantity is quantized, it is possible to uniformly quantize it, but when a separate manual adjustment means is provided, control instructions for that adjuster are given in stages. Quantization can also be measured non-linearly depending on the relationship between the quantity (for example, the rotation angle of the volume) and the control level. CPU (Central Processing Unit) 1 consisting of a microcomputer, etc.
8 is connected to another standard pattern memory 3, in which voice commands (control commands) of the commanding person are stored in advance together with a code specifying the control contents in a sampled and quantized form.Control commands To register voice (voice command) to standard pattern memory,
For example, in the case of controlling a television receiver, it is performed as follows. FIG. 3 shows an example of a control panel of a television receiver, which includes an input microphone 20, a registration mode switch 21, a dispatcher (speaker) number designation switch 1, 22-1, 222-2, etc. selected by the dispatcher, control Command designation switch "Power" designation switch 23 "Volume" designation switch 24 corresponding to power on/off switching, volume change, channel switching, respectively
Numerical buttons 26- are used as "channel" designation switch 25 and also to designate volume and channel.
1, 26-2, 26-3...26-11, 26-1
2 corresponds to each display lamp 27-1, 27-2,
...It is arranged together with 27-12. Further, the "OK" indicator lamp 28 at the bottom lights up when recognition or registration is successfully completed, and the "REPEAT" indicator lamp 29 lights up when the same goes wrong. To register a standard pattern using the registration control unit 30 , first press the registration switch 21 to enter the registration mode, then specify the speaker number with the specification switch 22-1 or 22-2, and then press the registration switch 21 to enter the registration mode. , and then press the "power" switch 23 one after another to select, for example, "DENGE" or "POWER".
Press the "volume" switch 24 and say "only" or "volume". When the "channel" switch 25 is pressed, the registration control circuit (not shown) outputs a mode switching signal and switches the output of the channel pushbutton switch circuit 31 via the switching circuit 32 to the registration control circuit side. As a result, when the numerical designation buttons 26-1, 26-2... (FIG. 3) included in the switch circuit 31 are pressed and ``1'', ``ni'', etc. are uttered, the respective sounds are transmitted to the input section 1 and the feature extraction section 2. Each control content (power supply,
are stored in the standard pattern memory 3 along with codes corresponding to the volume channels 1, 2, 3, . . . ).

Now, in the normal recognition mode, the aforementioned control voice is input, and the feature extraction filters 13-1, 13-2...
The signal string extracted and digitized by 13-N is stored in a storage memory 19 such as RAM, and then
The CPU 18 specifies the input voice by calculating the difference between this stored pattern and the standard pattern for all standard patterns and determining the standard pattern with the smallest difference. In general, even if human speech is uttered in the same language, the time course of the speech is not always the same, so some kind of time axis adjustment circuit as shown in FIG. 1 must be added. In FIG. 2, the time axis adjustment circuit is omitted for convenience of explanation.

In the recognition mode, voice is always captured, and when the input voice is interrupted, that is, during a pause period, the above-mentioned recognition calculation is executed, and the previous input voice is identified by pattern matching.
At this time, when it became possible to identify the input audio,
That is, when the input audio matches some standard pattern within an allowable error range, the CPU 18 outputs an instruction to the output control circuit 33 to control each corresponding control element of the television receiver. For example, when the input voice "Dengen Iri (Kiri)" is recognized, the output control circuit 33 controls the power supply circuit 34 of the television receiver to turn on and off. Also, “Channel・
When the input voice ``**'' (** is a number from 1 to 12) is recognized, the output control circuit 33 outputs the output to the channel switching circuit 35, thereby switching and controlling the tuner.

As mentioned above, the acquisition of voice signals in the voice command recognition mode is always performed while the command continues, and recognition calculations are executed by the CPU 18 when the input voice is interrupted, that is, at intervals of a certain command unit (pause period). The input voice commands up to that point are identified by the pattern matching method.

As mentioned above, in order to avoid malfunctions caused by the speaker output of the TV receiver, which is the controlled device, and other similar sounds issued by persons other than the commanding person, the comparative tolerance of pattern matching is set to be somewhat large for the most severe command word. At this time, when it becomes possible to specify the input voice, that is, when the input voice matches some standard pattern within an allowable error range,
The CPU 18 controls the output control circuit 33 to mute the audio output of the television receiver for a certain period of time. In the case of FIG. 2, the output control circuit 33 is connected to the collector of the transistor via the capacitor 58 by lowering the voltage Vc of the bias circuit 61 of the output amplification transistor 57 of the audio demodulation amplification circuit 56 of the television receiver. The audio output from the speaker 59 is stopped. Note that muting is not necessary for the earphone circuit 60 connected to the output side of the audio circuit 56. Normal control command words are, for example, "dengen", "iri", "dengen", "kiri", "channel", "ichi", "channel",
For example, when the input voice ``channel'' is recognized, the speaker voice is muted, and the subsequent voice input of ``ichi'' or ``ni'' is transmitted to the TV receiver. Since the sound generated by the noise is eliminated, the signal-to-noise ratio is greatly improved, and the recognition rate is greatly improved. This muting of audio does not result in any functional drawbacks, regardless of the control contents, such as changing the power on/off channel or changing the volume.

In the above embodiment, a method was adopted in which the bias voltage of the audio output transistor was grounded, but instead, a method of changing the gain of the drive stage,
Alternatively, a method may be adopted in which an attenuation circuit is inserted only during muting at a location unrelated to the earphone circuit, and in this case, the characteristics of the attenuation circuit include the filters 13-1, 13-2...13 of the feature extracting section 2 . −
A filter with characteristics opposite to N multiple filters is
Needless to say, a configuration may be adopted in which the signal is connected in series to the signal path in the audio circuit only during temporary muting via a switching element that is on/off controlled by the CPU output. Moreover, if a digital filter is used as such a filter, it is also possible to implement it into an IC. Since the present invention has the above-described configuration, by controlling the audio output circuit of an audio device such as a television receiver equipped with a voice recognition device with the output of the output control circuit of the voice recognition device, when input voice is sensed, By attenuating the output audio of the audio equipment to an appropriate level, the subsequent input audio can be improved.
This increases the SN ratio and is therefore extremely effective in improving the recognition rate of this type of speech recognition device.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a main part of a speech recognition device, FIG. 2 is a circuit for implementing a main part of the present invention, and FIG. 3 is a front view of an operation panel of a controlled device. 1 ...Input unit, 2 ...Feature extraction unit, 3 ...Standard pattern memory, 18...CPU, 37, 41, 45, 54,
68...AND gate, 44, 46...OR gate, 42, 50...Latch circuit, 53...Analog switch, 60...Voltage dividing resistor, 43, 69...
A-D conversion circuit, 57...Audio output transistor, 59...Speaker, 61...Bias circuit.

Claims (1)

[Claims] 1. An input unit that converts a first input voice and a second input voice that follows the first input voice into electrical signals, a feature extraction unit that extracts features of the electrical signal, and a standard pattern storage unit that stores voice features as a standard pattern; and a recognition processing unit that compares the feature pattern extracted by the feature extraction unit and the standard pattern to identify the first and second input voices. , an output control section that controls a television receiver or the like having a sound emitting means based on the recognition result of the recognition processing section;
and a muting circuit that mutates the audio output from the sound emitting means in response to a signal from the output control section for a period during which the input section inputs the second input audio. Control device by.