вход по аккаунту



код для вставкиСкачать
Patent Translate
Powered by EPO and Google
This translation is machine-generated. It cannot be guaranteed that it is intelligible, accurate,
complete, reliable or fit for specific purposes. Critical decisions, such as commercially relevant or
financial decisions, should not be based on machine-translation output.
The present invention relates to a speech input device, more particularly to a speech recognition
device in a speaker verification device and a speech recognition device, for example, keywords
for use in a speaker verification device. Also as a voice input means when you do not want people
to hear. When used in a voice input device, it is suitable to be used as a voice input means when
the contents being input are not to be heard by people around. Recently, attention has been
focused on speaker verification technology as an alternative to the conventional key in security
systems of various types of security systems, and research and development are actively
conducted. This is to compare the voice feature of the uttered keyword with the voice feature of
the specific speaker's keyword registered in advance, and determine whether or not it is the same
speaker's voice. However, with the present speaker verification technology, when another person
utters the same keyword as the keyword of a specific speaker registered in advance, the speaker
itself can not always be accurately checked, so the keyword itself is known to others. It is
important not to However, in the usual voice input method, keywords uttered to the microphone
can be easily heard by others around them, which is problematic. Moreover, even in various voice
recognition devices, in the normal voice input method, the voice uttered toward the microphone
can be heard by others around, so there is a problem when the content to be input into the
device is not to be known to the surrounding people. is there. One possible solution to this
problem is to squeeze out the input voice by uttering noise around the microphone. In this case,
the generated noise is also input from the microphone along with the voice, so the speaker
verification device or voice This is not preferable because the performance of a speech
processing device such as a recognition device is significantly reduced. {Circle over (1)} Once the
present invention has been made in view of the circumstances as described above, in particular,
keywords in speaker verification and the like. It is an object of the present invention to provide a
voice input device which can input voices of confidential contents in voice recognition to these
voice processing devices without being heard by the surrounding people. In order to achieve the
objects described in the present invention, a plurality of n noise generators that generate a
plurality of n noises correlated with each other and a plurality of the n pieces of noise are
respectively reproduced in advance. A plurality of n speakers respectively arranged in a
predetermined positional relationship, and a microphone disposed in a predetermined positional
relationship with the plurality of speakers for inputting voice, the plurality n of the plurality The
noise generator is characterized in that noise reproduced from the plurality of n speakers
generates noise which is mutually offset at the position of the microphone.
Hereinafter, the present invention will be described based on examples of the present invention.
FIG. 1 is a configuration diagram for explaining an embodiment of the present invention, in
which 1 is a first speaker, 2 is a second speaker, 3 is a first noise generator, and 4 is a second.
Noise generator, 5 is a speaker, 6 is a microphone, and the first speaker 1 and the second
speaker 2 are placed on the same wall surface, in order to prevent unnecessary reflected sound.
is there. The distance between the first speaker 1 and the micro is r, and the distance between
the second speaker 2 and the micro is r− (<rl). The first speaker 1 and the second speaker 2
have the same characteristics. The first noise generator 3 generates noise to be reproduced from
the first speaker 1, and the second noise generator 4 generates noise to be reproduced from the
second speaker 2. The micro converts the speech produced by the speaker 5 into an electrical
signal. This electrical signal is sent to the voice processing unit to be connected to the
subsequent stage. FIG. 2 is a diagram showing the details of the first noise generator 3 and the
second noise generator 4, wherein the first noise generator 3 comprises an oscillator 30, a first
adder 32, and a first noise generator 3. The oscillator 30 generates m sine waves, Xt (t) = atcos
(.omega.ut + at) (i = LLwsm) (1). The phases of these m sine waves are independent of one
another. The first adder 32 combines the m sine waves generated by the oscillator 30. とする。
The first power amplifier 34 amplifies the waveform synthesized by the first adder 32 and sends
it to the first speaker 1 as a noise waveform. The second noise generator 4 includes a phase
shifter 41, a second adder 42. The phase shifter 41 comprises a level adjuster 43 and a second
power amplifier 44. The m sine waves generated by the oscillator 30 are each delayed by a
predetermined phase (θ 1) by m sets of delay elements each having a predetermined delay time
(OI / ω). That is, the signal obtained from the phase shifter 41 is. X4 (t) = a ICO8 (6) st + (11t)
(3). Here, α ′ i = α 1 8 m (4), θ is x 1. It is # s = Ki (r, -rx) +2 ng (n = integer) as a wave
number with respect to (omega) 1 (= 2 (pi) / (lambda), (lambda) l 2 wavelength). The second
adder 42 synthesizes m sine waves of the output of the phase shifter 41, and the level adjuster
43 further performs level division by resistance division of the waveform synthesized by the
second adder 42. Double (r, / r,).
It is assumed that yx (t) = (rslrJ.yf (t) =. [email protected] (.omega.1t + a1) i = 1. Here, a'i = Skiru 0 r
Snow / Im (a). The second power amplifier 44 amplifies the output signal of the level adjuster 43
and sends it to the second speaker. The characteristics of the second power amplifier 44 are
equal to the characteristics of the first power amplifier 34. If the size of the first speaker 1 and
the second speaker 2 is sufficiently small with respect to the distances r% and r to the respective
microphones, it is considered that the noise reproduced from both speakers is transmitted as a
spherical wave Well, at the micro location, these noises will be layer A1. ここで、
A*/At=at/at=rl、rx(to)である。 The first term of equation (9) is due to yt
(t) of the first loudspeaker 1 and the second term is due to yx (t) of the second loudspeaker 2.
Here, when each angular velocity ω, that is, each wave number is divided into one and
calculated, equations (4), (5) and (10) are used. =? Jexp (j ( + at ktrJ) eXP (j (
+ al + 61 htr-)) 噌 [eXP (j ( + at ktrl))-eXP (j (.omega. + T + c ++ 2nz ktr-))] = 0, and
equation (9) is f (T) = O (12). Therefore, the noise reproduced by the first speaker 1 and the noise
reproduced by the second speaker 2. It is offset at the micro position and does not disturb the
micro voice input. In the present embodiment, the number of speakers is two, but a combination
of similar speakers or It can also be realized by three or more loudspeakers by using a
combination of loudspeakers such that the noise f (t) at the micro position is O. また。 By making
the delay time of the m delay elements of the phase shifter 41 of this embodiment variable. It is
also easy to adaptively control a signal reproduced from the second speaker 2 according to a
change in sound velocity or the like by using a microphone input when there is no voice input. As
is apparent from the description above, according to the present invention, noise is generated
from a plurality of speakers and voice input is performed using microphones in the vicinity
thereof. It is possible to input into these speech processing devices without having a human
being hear a voice of a certain content. Furthermore, since the above-mentioned plurality of
speakers generate noise which is canceled only at a limited position, and the above-mentioned
microphone is arranged at that position, the recognition performance of the above-mentioned
voice processing device is not impaired. There are advantages such as
Brief description of the drawings
FIG. 1 is a configuration diagram for explaining an embodiment of the present invention, and FIG.
2 is a diagram for explaining the noise generator shown in FIG. 1 and the details thereof.
1 ... 1st speaker, 2 ... 2nd speaker, 3 ... 1st noise generator, 4 ... 2nd noise generator, 5 ... speaker,
6 ... -Mike.
Без категории
Размер файла
12 Кб
description, jph03149599
Пожаловаться на содержимое документа