KR101615766B1 - Impulsive noise detector, method of detecting impulsive noise and impulsive noise remover system - Google Patents

Abstract

The present invention relates to a frequency converter for dividing input digital signals into m blocks, converting the digital signals into frequency domain in units of the divided blocks, and outputting m frequency domain signals converted into a frequency domain, A first energy calculator for calculating m energy values existing in a high frequency band for each of the m frequency domain signals input from the frequency converter and outputting m first energy values, A first background signal energy calculator for estimating m first background signal energy values using the first energy values and m first energy values and m first background signal energy values, Comparing the first energy values and the first background signal energy values to detect first impulsive noises, The present invention relates to unexpected noise detector comprises a first comparator for outputting the first noise signal a sudden 1 abrupt noise are included.

Description

TECHNICAL FIELD [0001] The present invention relates to an abrupt noise detector, an abrupt noise detection method, and a sudden noise removal system.

The present invention relates to a sudden noise detector and a method capable of accurately detecting a position of the noise when a voice signal including various noises generated in the surrounding environment is received.

In an environment for receiving a voice signal, various noise may exist in addition to a voice signal. For example, when a place where a voice signal is received is in a station where a lot of people are located, an airport, a bus, a subway, a street, or the like, various noise other than the voice signal to be received exists. Therefore, if a voice signal is received at the above-mentioned location, the received voice signal contains various noises. Such noise includes a sound of closing a desk drawer, a sound of closing a door, and a sound of clapping. The noise may exist for a short period of time and may be a very large and simple waveform, a waveform that exists for a relatively long period of time and has a complex amplitude and signal component similar to speech.

Since speech signals including various noise signals as described above are often received, studies are being conducted on techniques for removing noise from the speech signal. More specifically, a method for detecting noise from the voice signal, a noise canceling apparatus for removing the detected noise, and the like have been actively studied.

However, there is a problem that it is difficult to distinguish between a noise and a voice signal when detecting using a pitch of a voice signal in the process of detecting a noise, so that accurate detection is difficult. Also, when a noise occurs with a large size and a relatively small size occurs at the end of the noise, a noise having a relatively small size is almost the same as a voice signal.

SUMMARY OF THE INVENTION It is an object of the present invention to overcome the above-described problems by providing a detector and method for accurately detecting noise in a voice signal including various kinds of noise.

The noise detector according to the present invention is a noise detector that divides input digital signals into m blocks, converts the digital signals into frequency domains in units of the divided blocks, and outputs m frequency domain signals converted into a frequency domain A first energy calculator for calculating energy in a high frequency band for each of m frequency domain signals input from the frequency converter and outputting m first energy values; A first background signal energy calculator for estimating m first background signal energy values using m first energy values input from the energy calculator, and a second background energy calculator for calculating m first energy values And when m first background signal energy values are mapped from the first background signal energy calculation unit And a first comparator for comparing the first energy values and the first background signal energy values to detect first burst noise and outputting a first burst noise signal including the first burst noise, .

Accordingly, the sudden noise detector according to the present invention can detect the sudden noise clearly even when the pitch of the sudden noise is similar to the pitch of the voice signal.

A second energy calculator for calculating a second energy value based on each of the digital signals input from the analog digital converter and outputting the second energy value; A second background signal energy calculator for estimating second background signal energy values using the second energy values and outputting the estimated second background signal energy values; Values and the second background signal energy values input from the second background signal energy operation unit, comparing the second energy values and the second background signal energy values to detect second sudden noise, A second comparator for outputting a second burst noise signal including second burst noise and a second comparator for outputting the second burst noise signal detected as m And outputting a third sudden noise signal reflecting the calculated result, and outputting a third sudden noise signal reflecting the calculated result when the number of the second sudden noise existing in each block is equal to or greater than a predetermined number, And a sudden noise detector.

Accordingly, the sudden noise detector according to the present invention can accurately detect a sudden noise even when a voice signal having a high frequency component is input and when the peak of the sudden noise is similar to the peak of the voice signal.

A method for detecting an artificial noise according to the present invention comprises the steps of dividing input digital signals into m blocks, converting the digital signals into a frequency domain in units of the divided blocks, and outputting m frequency domain signals converted into a frequency domain A first energy value calculation step of calculating m first energy values by calculating an energy existing in a high frequency band of the m frequency domain signals, and calculating m first energy values using m first energy values, A first background signal energy value calculation step of estimating first background signal energy values and a second background energy value calculation step of calculating m first energy values calculated in the first energy value calculation step and m Corresponding to the first background signal energy values, and comparing the first energy values and the first background signal energy values Detecting a first sudden noise, a first sudden noise detecting step for outputting a first abrupt noise signal with the noise they broke.

Accordingly, even when the pitch of the sudden noise is similar to the pitch of the voice signal, the sudden noise detection method according to the present invention can clearly detect the sudden noise.

The sudden noise elimination system according to the present invention comprises: the sudden noise detector; And a signal analyzer configured to perform a long term prediction (LTP) analysis on the input digital signals, a periodic signal generator for generating a periodic signal based on the analysis result, An LTP analyzing unit including a subtractor configured to subtract an output signal from the LTP analyzer based on an abrupt noise signal output from the abrupt noise detecting unit and an abrupt noise attenuator configured to attenuate abrupt noise in a signal output from the LTP analyzing unit, And a summation unit configured to sum the output signal and the periodic signal output from the periodic signal generating unit, wherein the second energy calculator comprises: a second energy calculator configured to calculate a second energy based on the signal output from the LTP analyzer, And calculates an energy value.

According to the present invention, even if the peak value of the sudden noise is similar to the peak value of the voice signal, there is an advantage that the sudden noise can be accurately detected.

Figs. 1A to 1C are diagrams showing a configuration and a detection method according to an embodiment of a detector for detecting a sudden noise.

1A, the detector includes a receiver 100, an analog-to-digital converter (ADC) 110, a sudden noise amplifier 120, a second energy calculator 130, a second background signal energy calculator 140, 2 < / RTI >

The receiving unit 100 is a device capable of receiving a voice signal and transmits the received voice signal to the analog-to-digital converter (ADC)

The analog-to-digital converter (ADC) 110 converts the speech signal input from the receiver 100 into n digital signals and transmits the n digital signals to the disturbance noise amplifier 120. Referring to FIG. 1B, the analog-to-digital converter 110 converts the input voice signal into n digital signals.

The sudden noise amplifier 120 amplifies n digital signals input from the analog-to-digital converter 110. The amplification uses a quadratic differentiator, and the output of the sudden noise amplifier 120 is a quadratic differential coefficient. The sudden noise amplifier 120 may further enhance the characteristic of the sudden noise by amplifying the sudden noise. That is, the magnitude of the sudden noise becomes larger than the magnitude of the sudden noise in the past.

For example, z (n) = x (n-1) - 2 * x (n) + x (n + 1) can be used. N (n) denotes the output of the noise amplifier, x (n) denotes the output of the analog-to-digital converter 110 and n denotes the order of the digital signal (0, 1, 2, n).

The sudden noise amplifier 120 transmits the output z (n) to the second energy calculator 130.

The second energy calculator 130 calculates the second energy w (n) using the output z (n) input from the sudden noise amplifier 120. The operation of w (n) is calculated using the following equation.

w (n): the output of the second energy calculator

M: Number of digital signals used in the window

z (n): the output of the sudden noise amplifier

For example, when M = 10, j changes from -5 to 5, and n changes from 0 to n. (N) (second energy value) is calculated by sequentially adding the values of the variables. The number of w (n) is output. The values and equations of the variables are only examples, and the user can use various modifications as needed.

The second energy calculator 130 transmits the second energy values w (n) to the second background signal energy calculator 140 and the second comparator 150.

The second background signal energy calculator 140 estimates the second background signal energy using the second energy values input from the second energy calculator 130. The background signal refers to a signal excluding a sudden noise in a voice signal. The background signal energy refers to a value obtained by calculating the energy of the background signals, which are signals excluding the sudden noise in the speech signal. In general, since it is difficult to obtain only the background signal, the background signal energy value is estimated. The second background signal energy calculator 140 calculates a recursive median filter, an intermediate value filter, an average value filter, and the like. The intermediate value means that among the values existing in the filter, the rank of the magnitude outputs values belonging to the intermediate rank. The average value means that an average value of the values existing in the filter is calculated to output an average value. A method of calculating the average value by adding a weight value can also be used. The recursive column refers to a case where an output of a filter obtained in the past is also used during the input of the filter.

When the second background signal energy calculator 140 uses a recursive intermediate value filter, it is calculated using the following equation.

b (n): second background signal energy value

w (n): second energy value

k: the size of the recursive median filter

Med represents an intermediate value, and RMF represents a recursive intermediate value filter. b (n-k), ..., b (n-1) denotes a result value of the first background signal energy values obtained in the past through the first energy calculator 170. For example, if k is 10 and w (10) is input, the second background signal energy calculator 140 calculates b (0), b (1), b (10), w (11), w (12), ..., w (20). If the value of w (11) corresponds to an intermediate value, the second background signal energy calculator 140 outputs w (11) as a second background signal energy value when w (10) is input. In this manner, the second background signal energy calculator 140 outputs n second background signal energy values.

The second comparator 150 compares n second energy values input from the second energy calculator 130 and n input second background signal energy values from the second energy background signal calculator 140, . The second comparator 150 compares the corresponding second energy values and second background signal energy values to detect the position of the sudden noise.

For example, the second comparator 150 may detect a case where the second energy value is greater than a corresponding second background signal energy value as a sudden noise. If the second energy value of the third digital signal is 10 and the second background signal energy value is 8, the second comparator 150 detects that there is a second sudden noise at the position of the third digital signal. By repeating the above process, the second comparator 150 detects the position of the second incident noise and outputs the second incident noise signal including the second incident noise. The second unexpected noise refers to a noise detected as a sudden noise in the second comparator 150. The second nasal noise signal includes the second nasal noise, and the second nasal noise signal indicates a position of the second nasal noise.

In addition, the second comparator 150 may detect a case where the second energy value is greater than a predetermined value, which is greater than the corresponding second background signal energy value, as a sudden noise. The predetermined value is arbitrarily set by the user. For example, if the user is set to detect as sudden noise only if the second energy value and the second background signal energy value are greater than or equal to 3, the second energy value of the third digital signal is 10, If the value is 8, the third digital signal is not detected as a sudden noise.

In addition, the second comparator 150 may detect a case where the ratio of the second background energy value and the second background energy value is greater than a preset value, as the sudden noise. For example, it can be calculated using the following equation.

               d (n) = (w (n) - b (n)) / b (n)

               g (n) = 1, d (n) > C

                    = 0 otherwise

w (n): second energy values

b (n): second background signal energy values

d (n): the ratio of the second energy value and the second background signal energy value

g (n): the second burst noise signal

C: Constant constant

n: order of digital signals (0, 1, 2, ..., n)

If the user is set to detect the ratio C of 0.1 or more as a sudden noise, if the second energy value of the third digital signal is 10 and the second background signal energy value is 8, then (10 - 8) / 8 = 1/4 = 0.25, so it is detected as a sudden noise.

In addition to the above-described method, the user can set a criterion for detecting whether the noise is sudden noise in the second comparator 150 by various methods.

Referring to FIG. 1C, an exemplary second burst noise signal g (n) output from the second comparator 150 is shown. Referring to FIG. 1C, when n is 1, 4, 9, 11, 12, 13, .., it can be seen that a second nuisance noise occurs.

Accordingly, the detector according to the present embodiment can detect the position of the sudden noise. However, the detector uses a pitch of the voice to detect it, so that there is a problem that it can not be detected as a sudden noise when the pitch of the voice and the pitch of the sudden noise are similar. In addition, when the sudden noise is generated and the pitch of the sudden noise is relatively small at the end of the sudden noise, the detector can not detect the sudden noise.

FIGS. 2A to 2F are diagrams showing a configuration and a detection method according to an embodiment of a detector for detecting an unexpected noise according to the present invention.

FIG. 2A is a diagram illustrating the magnitude (dB) of energy according to a change in frequency when both sudden noises and voice signals are present and only voice signals are present when sudden noise is present.

Referring to FIG. 2A, it can be seen that when there are both sudden noise and sudden noise and voice signals, the energy value in the high frequency band is as high as 60 dB to 90 dB. On the other hand, when only voice signals are present, the energy value in the high frequency band is as small as 30 dB to 50 dB. Accordingly, in the present invention, after the speech signal including the sudden noise is converted into the frequency domain signal using the above characteristics, the sudden noise is detected by comparing the energy values. Hereinafter, a detailed detection method of the detector according to the present invention will be described.

2B, the detector according to the present invention includes a receiver 100, an analog-to-digital converter (ADC) 110, a frequency converter 160, a first energy calculator 170, a first background signal energy calculator 180 And a first comparator 190.

The receiving unit 100 is a device capable of receiving a voice signal and transmits the received voice signal to the analog-to-digital converter (ADC)

The analog-to-digital converter (ADC) 110 converts the speech signal input from the receiver 100 into n digital signals, and transmits the n digital signals thus converted to the frequency converter 160. Referring to FIG. 2C, the analog-to-digital converter 110 converts the input voice signal into n digital signals.

The frequency converter 160 divides n digital signals input from the analog-to-digital converter 110 into m blocks. The frequency converter 160 may use a Fourier transform or the like. Referring to FIG. 2C, n digital signals are divided into m blocks. The frequency conversion unit 160 And converts the digital signals into the frequency domain in the divided block units (m). That is, the frequency converter 160 converts the time-domain digital signals existing in m blocks into the frequency domain, and outputs m frequency domain signals.

The frequency converter 160 transmits m frequency domain signals to the first energy calculator 170.

The first energy calculator 170 calculates m first energy values using the m frequency-domain signals input from the frequency converter 160. The calculation of the first energy value can be obtained by calculating only energy values existing in the high frequency region of the frequency response. For example, the first energy value can be calculated by calculating only the energy existing in the frequency range of 2.5 kHz to 3.5 kHz. The user can freely set the frequency range from 1500Hz to 6000Hz. By repeating the above process, the first energy calculator 170 can calculate m first energy values. The first energy calculator 170 transmits the first energy values to the first background signal energy calculator 180 and the first comparator 190.

The first background signal energy calculator 180 estimates the first background signal energy using the first energy values input from the first energy calculator 170. The background signal refers to a signal excluding a sudden noise in a voice signal. The background signal energy refers to a value obtained by calculating the energy of the background signals, which are signals excluding the sudden noise in the speech signal. In general, since it is difficult to obtain only the background signal, the background signal energy value is estimated. The first background signal energy calculator 170 calculates a recursive median filter, an intermediate value filter, an average value filter, and the like. The intermediate value means that among the values existing in the filter, the rank of the magnitude outputs values belonging to the intermediate rank. The average value means that an average value of the values existing in the filter is calculated to output an average value. A method of calculating the average value by adding a weight value may also be used. The recursive column refers to a case where an output of a filter obtained in the past is also used during the input of the filter.

When the first background signal energy calculator 170 uses a recursive intermediate value filter, it is calculated using the following equation.

(m), h (m) = median {h (m-u), ..., h

h (m): first background signal energy value

e (m): first energy value

u: the size of the recursive median filter

Med represents an intermediate value, and RMF represents a recursive intermediate value filter. h (n-u), ..., h (n-1) denotes a result value of the first background signal energy values obtained in the past through the first energy calculator 170. For example, if u is 10 and e (10) is input, the first background signal energy calculator 170 calculates h (0), h (1), h And outputs an energy value belonging to an intermediate rank among the values of e (10), e (11), e (12), ..., e (20). If e (11) corresponds to a middle value, the first background signal energy calculator 170 outputs e (11) as a first background signal energy value when e (10) is input. In this manner, the first background signal energy calculator 170 outputs m second background signal energy values.

The first comparator 190 outputs m first energy values input from the first energy calculator 170 and m first background signal energy values from the first energy background signal calculator 180, . The first comparator 190 compares the first energy values and the first background energy values with each other to detect the location of the sudden noise.

For example, the first comparator 190 may detect a case where the first energy value is greater than a corresponding first background signal energy value as a sudden noise. When the first energy value of the third digital signal is 12 and the first background signal energy value is 8, the first comparator 190 detects that there is a first sudden noise at the position of the third digital signal. By repeating the above process, the first comparator 190 detects the position of the first incident noise and outputs the first incident noise signal including the first incident noise. The first sudden noise refers to a noise detected as a sudden noise in the first comparator 190. The first sudden noise signal is a signal including the first sudden noise, which indicates the position of the first sudden noise.

In addition, the first comparator 190 may detect a case where the first energy value is greater than a predetermined value greater than the corresponding first background signal energy value as a sudden noise. The predetermined value is arbitrarily set by the user. For example, if the user sets the first energy value to be equal to or greater than three times the first energy value and the first background energy value, the first energy value of the third digital signal is 12, When the energy value is 8, the third digital signal is detected as the first sudden noise.

Also, the first comparator 190 may detect a case where the ratio of the first background energy value and the first background energy value is greater than a predetermined value as a first incident noise. For example, it can be calculated using the following equation.

               p (m) = (e (m) - h (m)) / h (m)

               q (m) = 1, p (m) > C

                    = 0 otherwise

e (m): first energy values

h (m): first background signal energy values

p (m): a ratio of a first energy value and a second background signal energy value

q (m): first sudden noise signal

C: Constant constant

m: order of blocks (0, 1, 2, ..., m)

If the user is set to detect the ratio C of 0.3 or more as a sudden noise, if the first energy value of the third block is 10 and the first background signal energy value is 8, then (10 - 8) / 8 = 1/4 = 0.25, so it is not detected as sudden noise.

In addition to the above-described method, the user can set a criterion for detecting whether the noise is sudden noise in the second comparator 150 by various methods.

Referring to FIG. 2D, an example of the first incident noise signal g (n) output from the first comparator 190 is shown. Referring to FIG. 2D, when m is 1, 2, 4, ..., m-1, it can be seen that a first sudden noise has occurred.

FIG. 2E is a view showing a position where a sudden noise exists, and FIG. 2F is a diagram of detecting a sudden noise using a detector according to the present invention.

Referring to FIGS. 2E and 2F, it can be seen that almost all the sudden noise can be detected by using the detector according to the present invention. The vertical axis of FIG. 2E is a numerical value representing the presence of a sudden noise of the corresponding block by binarization, and the horizontal axis represents the order of blocks. That is, when the number of the horizontal axis is 200, it means the 200th block. The vertical axis in FIG. 2F indicates the energy value of the corresponding block, and the horizontal axis indicates the order of blocks.

Accordingly, the detector according to the present embodiment can detect the sudden noise clearly even when the pitch of the sudden noise is similar to the pitch of the voice signal. In addition, the detector according to the present invention can detect almost all the sudden noises accurately in a section where a vowel exists.

FIGS. 3A through 3E are diagrams illustrating a detector according to an exemplary embodiment of the present invention.

3A, a detector according to the present invention includes a receiver 100, an analog-to-digital converter (ADC) 110, a sudden noise amplifier 120, a second energy calculator 130, a second background signal energy calculator 140 A second comparator 150, a frequency converter 160, a first energy calculator 170, a first background signal energy calculator 180, a first comparator 190, and a sudden noise detector 200 .

1A through 2F except for the above-mentioned sudden noise detector 200, the description thereof will be omitted.

The sudden noise detector 200 divides the second burst noise signal g (n) output from the second comparator 150 into m blocks. The m blocks are equal to the number of blocks divided by the frequency transforming unit 160. Referring to FIG. 3B, the burst noise detector 200 divides the second burst noise signal into m blocks. The sudden noise detection unit 200 detects an unexpected noise only when the number of the second sudden noise in each block is equal to or greater than a predetermined value. For example, if the number of the second sudden noise is three or more, it is detected as a sudden noise. If m = 1, the sudden noise detecting unit 200 detects the second sudden noise as two It is not detected. On the other hand, when m = 2, the sudden noise detector 200 detects the sudden noise because the number of the second sudden noises is three. By repeating this process, the sudden noise detector 200 converts the second nasal noise signal into m third nasal noise signals. Referring to FIG. 3C, the third burst noise signal is shown. The number of the second sudden noises is three or more. However, the user can freely define the number of the second sudden noises to 10, 30, 100, 200, 500, and so on.

The sudden noise detector 200 may be configured to match the first and third unexpected noise signals input from the first comparator 190 and output the first and third sudden noise signals, It is calculated that there is a sudden noise only in the presence of noise. For example, referring to FIG. 3C and FIG. 3D, when m = 1, the third nuisance signal has a nuisance noise, but since the first nuisance signal has no nuisance noise, The detection unit 200 calculates that there is no final sudden noise. Also, when m = 2, since the third nuisance signal and the first nuisance noise signal have nuisance noise, the nuisance noise detector 200 calculates that there is a final nuisance noise. By repeating this process, the final sudden noise is detected. Referring to FIG. 3E, the result of the sudden noise detector 200 detecting the final sudden noise is shown in FIG. 3C and FIG. 3D.

In addition, the sudden noise detector 200 determines whether there is a sudden noise in a corresponding block of the third burst noise signal only when there is a first burst noise in the corresponding block of the first burst noise signal So that the final sudden noise can be detected. For example, referring to FIG. 3C and FIG. 3D, in the case where m = 1, the sudden noise detector 200 detects a first sudden noise in the first sudden noise signal, 1 < / RTI > Since the sudden noise detector 200 does not have a sudden noise in a portion corresponding to m = 1 of the third burst noise signal, it calculates that there is no final sudden noise in a block corresponding to m = 1. Since the first sudden noise signal does not exist in the first sudden noise signal when m = 3, the sudden noise detecting unit 200 does not need to determine whether there is a sudden noise of the third sudden noise signal, and m = 3 The final block noise is not present in the corresponding block. As described above, the sudden noise detector 200 determines whether there is a sudden noise in the block corresponding to the third sudden noise signal only when the sudden noise of the block corresponding to the first sudden noise signal exists. By doing so, the sudden noise detection unit 200 can reduce the amount of data that can be calculated.

In addition, the sudden noise detector 200 determines whether there is a sudden noise in a corresponding block of the first sudden noise signal only when there is a sudden noise in the corresponding block of the calculated second sudden noise signal It is possible.

Accordingly, when the sudden noise is detected using the energy of the high frequency band, there is a problem that if a strong voice signal having a high frequency component is input, it is erroneously recognized as a sudden noise. On the other hand, when a sudden noise is detected using a peak component of a signal, there is a problem that when the peak of the sudden noise is similar to the peak of a voice signal, it is not recognized as a sudden noise. Therefore, when the sudden noise detector according to the present invention is used, the sudden noise detection error can be remarkably reduced by detecting the sudden noise only in the case where all the sudden noise exists in the results of the two methods for detecting the sudden noise.

In the present embodiment, the second energy calculator 130, the second background signal energy calculator 140, the second comparator 150, the frequency converter 160, the first energy calculator 170, The first comparator 190, and the sudden noise detector 200 may be formed of one chip or may be formed of a plurality of chips according to functions.

4A and 4B are flowcharts of a method for detecting an artificial noise according to the present invention.

A voice signal is received through the receiving unit (S400). The analog-to-digital converter converts the received speech signal into n digital signals (S410). The Fourier transformer divides the n digital signals received from the analog-to-digital converter into m blocks, performs Fourier transform on each of the blocks, and performs Fourier transform to calculate m frequency responses at step S420. The first energy calculator calculates m first energy values by calculating energy in the high frequency band of the m frequency responses (S430). The first background signal energy value operation unit estimates m first background signal energy values using m first energy values input from the first energy operation unit (S440). The first background signal energy value calculator estimates the image signal using a recursive intermediate value filter, an intermediate value filter or an average value filter. The first comparator compares m first energy values and m first background signal energy values to detect first nuisance noise and outputs a first nuisance noise signal including the nuisance noise (S450) . The first incidental noise may be detected as a sudden noise when the first energy value is greater than the corresponding first background signal energy value. The first incident noise may be detected as a sudden noise when the first energy value is greater than a predetermined value greater than the corresponding first background signal energy value. Also, the detection of the first sudden noise may include a method of detecting a case where the ratio of the first energy value and the first background energy value is greater than a preset value as the sudden noise.

In operation S460, the second energy calculator computes n second energy values using the digital signals input before and after the predetermined number of the n digital signals input from the analog-to-digital converter. The second background signal energy calculator estimates n second background signal energy values using the n second energy values (S470). The second comparator compares n second energy values and n second background signal energy values to detect second nuisance noise, and outputs a second nuisance noise signal including the nuisance noise (S480) . The sudden noise detector divides the detected second sudden noise signal into m blocks (S490). The sudden noise detector calculates that there is a sudden noise only when the number of the second sudden noise in each block is equal to or greater than a predetermined number, and outputs a third sudden noise signal (S500). The sudden noise detector may be configured to match the third sudden noise signal with the first sudden noise signal detected by the first comparator and then to output only a final sudden noise only at a position where both the first sudden noise and the third sudden noise exist (Step S510). In step S510, a final sudden noise signal is detected. The detection of the final break noise is performed only when there is a first break noise of the first break noise signal and whether or not there is a second break noise of the second break noise signal, A method of calculating whether or not to use it can be used. In addition, the detection of the final sudden noise may be performed only when the second sudden noise of the second sudden noise signal is present, and it is determined whether or not the second sudden noise of the second sudden noise signal exists, A method of calculating whether or not to perform the operation can be used.

Accordingly, in the method of detecting an unexpected noise according to the present invention, it is possible to remarkably reduce an unexpected noise detection error by detecting only a final occurrence noise in the case where all the sudden noise exists in the results of the two methods for detecting the sudden noise.

5 is a block diagram of an abrupt noise reduction system in accordance with the present invention.

Referring to FIG. 5, the sudden noise elimination system includes a sudden noise detector 500 and a sudden noise remover 600.

3A to 3E, the description of the sudden noise detector 500 will be omitted.

The sudden noise canceller 600 includes an LTP (Long Term Predition) analyzer 610, a sudden noise attenuator 620, and a summer 630.

The LTP analyzer 610 performs a long term prediction (LTP) analysis based on the digital signals input from the analog-to-digital converter 110. The long term pre- The analysis can be used to determine pitch gain and pitch lag parameters. The periodicity of the input voice signal can be determined based on the parameter. Then, the LTP analyzer 610 generates a periodic signal based on the analysis result, and subtracts the periodic signal from the digital signal input from the analog-digital converter 110. The periodic signal will mostly correspond to a voiced sound. In addition, the signal obtained by subtracting the periodic signal from the input digital signal may include unvoiced and sudden noise. The signal output from the LTP analyzing unit 610 is transmitted to the sudden noise amplifier 120, the sudden noise attenuator 620, and the summer 630. A detailed description thereof will be made with reference to Fig. 6 below.

The sudden noise attenuator 620 attenuates the sudden noise among the signals output from the LTP analyzing unit 610 based on the sudden noise signal output from the sudden noise detecting unit 200. Specifically, the sudden noise attenuator 620 outputs a signal output from the LTP analyzer 610 corresponding to a position determined to have generated a sudden noise based on the sudden noise signal output from the sudden noise detector 200, As shown in FIG. Then, the sudden noise attenuator 620 attenuates the impulsive noise included in the detected signal and outputs the attenuated noise. Thus, the sudden noise can be removed.

The summer 630 sums and outputs the signal output from the sudden noise reducer 620 and the periodic signal. Specifically, the signal output from the sudden noise attenuator 620 includes unvoiced sound, and the periodic signal includes a sound signal. Accordingly, the signal output from the summer 630 may correspond to the voiced and unvoiced sounds of the unexpected noise. That is, the received voice signal corresponds to a signal from which the sudden noise is removed.

A control unit (not shown) controls the overall operation of the sudden noise reduction system.

Accordingly, it is possible to detect the sudden noise through the sudden noise detector, and the sudden noise canceler removes the sudden noise among the received voice signals based on the detected sudden noise information, It can be prevented from being received or transmitted together.

FIG. 6 is a block diagram embodying the LTP analysis unit of FIG. 5; FIG.

The LTP analyzer 610 includes a signal analyzer 611, a periodic signal generator 612, and a subtractor 613.

The signal analyzer 611 performs a long term prediction (LTP) analysis on the digital signals input from the analog-digital converter 110.

The periodic signal generator 612 generates a periodic signal based on the analyzed result of the signal analyzer 611 and outputs the periodic signal. The periodic signal is transmitted to the summer 630 and the subtractor 613. The signal output from the periodic signal generator 612 includes voiced sounds.

The subtracter 613 subtracts the periodic signal generated in the periodic signal generator 612 from the digital signals input from the analog-digital converter 110. [ The output signal of the subtractor 613 is transmitted to the sudden noise attenuator 620 and the noise reduction amplifier 120. The signal output from the subtractor 613 includes a sudden noise and an unvoiced sound.

FIG. 7 is a flowchart for explaining a method for eliminating the sudden noise in the sudden noise removing system according to the present invention.

5 and 7, a voice signal is received or transmitted through a wireless communication unit (not shown) or a microphone (not shown) (S700). Next, the control unit determines whether a sudden noise generation signal is input (S710). If a sudden noise generation signal is not input, the controller determines whether there is a sudden noise based on a signal input from the sudden noise detector 500 (S720). On the other hand, if a sudden noise generation signal is input, the control unit executes the following step S730. In step S720, if there is a sudden noise, the controller drives the sudden noise canceller 600 to remove the sudden noise from the voice signal received or transmitted (S730). The method of removing the sudden noise is as described in FIGS. 5 and 6. FIG. Then, the control unit transmits or receives the voice signal from which the sudden noise has been removed. Alternatively, the control unit records the voice signal according to the function.

Thus, according to the present invention, sudden noise can be removed from a voice signal.

In addition, according to the present invention, when a sudden noise generation signal such as a keypad selection signal is inputted, the sudden noise can be efficiently removed by driving the sudden noise eliminator immediately without driving the sudden noise detector to remove the sudden noise.

It is to be noted that the technical idea of the present invention has been specifically described in accordance with the above preferred embodiment, but the above-mentioned embodiments are intended to be illustrative and not restrictive. It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit and scope of the invention.

Figs. 1A to 1C are diagrams showing a configuration and a detection method according to an embodiment of a detector for detecting a sudden noise. Fig.

FIGS. 2A to 2F are diagrams showing a configuration and a detection method according to an embodiment of a detector for detecting an unexpected noise according to the present invention; FIG.

Figures 3A-3E illustrate a configuration according to an embodiment of an extinction noise detector in accordance with the present invention.

4 is a flow chart of a method for detecting an impulsive noise according to the present invention.

5 is a block diagram of a sudden noise reduction system in accordance with the present invention;

FIG. 6 is a block diagram embodying the LTP analysis unit of FIG. 5; FIG.

FIG. 7 is a flowchart for explaining a method for eliminating sudden noise in a sudden noise removing system according to the present invention. FIG.

Claims (18)

A frequency converter for dividing the input digital signals into m blocks, converting the digital signals into the frequency domain in units of the divided blocks, and outputting the m frequency domain signals converted into the frequency domain; A first energy calculator for calculating m energy values existing in a high frequency band for each of the m frequency domain signals input from the frequency converter; A first background signal energy calculator for estimating m first background signal energy values using m first energy values input from the first energy calculator; M first energy values from the first energy calculator and m first background energy values from the first background energy calculator and outputs the first energy values and the first background energy And a first comparator for comparing the energy values to detect first nuisance noises and outputting a first nuisance signal including the first nuisance noises. The method according to claim 1, A second energy calculator for calculating a second energy value based on each of the input digital signals and outputting the second energy value; A second background signal energy calculator for estimating second background signal energy values using the second energy values input from the second energy calculator and outputting the estimated second background signal energy values; The second energy values input from the second energy operation unit and the second background signal energy values input from the second background signal energy operation unit are correlated, and the second energy values and the second background signal energy value A second comparator for detecting second unintentional noises and outputting a second unintentional noise signal including the second unintentional noises; Dividing the detected second incident noise signal into m blocks and calculating that there is a sudden noise only when the number of the second incidental noise in each block is equal to or greater than a predetermined number, Further comprising a sudden noise detector for outputting a third sudden noise signal that reflects the third sudden noise signal. 3. The method of claim 2, Wherein the sudden noise detector comprises: And a third comparator for comparing the third incident noise signal with the first incident noise signal inputted from the first comparator and then comparing the third incident noise signal and the third incident noise signal, And outputs a final spurious noise signal reflecting the calculated result. 3. The method of claim 2, Wherein the sudden noise detector comprises: And determining whether there is a second disturbance noise of the second disturbance signal only if there is a first disturbance noise of the first incident noise signal, A sudden noise detector. 3. The method of claim 2, Wherein the sudden noise detector comprises: It is determined whether there is a second disturbance noise of the second disturbance noise signal only when there is a second disturbance noise of the second incident noise signal, A sudden noise detector. The method according to claim 1, Wherein the first background signal energy calculation unit comprises: Wherein the background signal energy value is estimated using at least one of an intermediate value filter, a recursive intermediate value filter, and an average value filter. The method according to claim 1, Wherein the first comparator comprises: When the first energy value is greater than the corresponding first background signal energy value and when the first energy value is greater than a predetermined value than the corresponding first background signal energy value, And the first background noise is detected based on any one of a case where the ratio of the first background signal energy value is larger than a predetermined value. The method according to claim 1, A receiver for receiving a voice signal; Further comprising an analog-to-digital converter for converting the voice signal into digital signals and outputting the converted digital signals to the frequency conversion unit. 3. The method of claim 2, Wherein the second background signal energy calculation unit comprises: Wherein the second background signal energy value is estimated using at least one of an intermediate value filter, a recursive intermediate value filter, and an average value filter. 3. The method of claim 2, Wherein the second comparator comprises: And detects a case where the second energy value is greater than the corresponding second background signal energy value as a sudden noise. A frequency conversion step of dividing input digital signals into m blocks, converting digital signals into frequency domain in the divided block units, and outputting m frequency domain signals converted into a frequency domain; A first energy value calculation step of calculating m first energy values by calculating an energy existing in a high frequency band of the m frequency domain signals; A first background signal energy value calculation step of estimating m first background signal energy values using the m first energy values; M first energy values calculated in the first energy value calculation step and m first background signal energy values calculated in the first background energy calculation step, And a first burst noise detection step of comparing the first background signal energy values to detect first burst noise, and outputting a first burst noise signal including the burst noise. 12. The method of claim 11, A second energy value calculation step of calculating a second energy value using the digital signal input in the frequency conversion step; A second background signal energy calculation step of estimating second background signal energy values using the second energy values; The second energy values calculated in the second energy value calculation step and the second background energy values calculated in the second energy value calculation step are compared to detect second sudden noises, A second disturbance noise detecting step of outputting a second disturbance noise signal Dividing the detected second incident noise signal into m blocks and calculating that there is a sudden noise only when the number of the second sudden noise in each block is equal to or greater than a predetermined number, And outputting a third burst noise signal outputting a signal. 13. The method of claim 12, The third incident noise signal and the first incident noise signal detected by the first comparator are matched with each other and then a final sudden noise exists only at a position where both the first incident noise and the third incident noise exist And outputting a final nuisance signal reflecting the result of the calculation. 14. The method of claim 13, Wherein the final burst noise detection step comprises: And determining whether there is a second disturbance noise of the second disturbance signal only if there is a first disturbance noise of the first incident noise signal, The method comprising the steps of: 14. The method of claim 13, Wherein the final burst noise detection step comprises: It is determined whether there is a second disturbance noise of the second disturbance noise signal only when there is a second disturbance noise of the second incident noise signal, The method comprising the steps of: A signal analyzer configured to perform a long term prediction (LTP) analysis on input digital signals, a periodic signal generator configured to generate and output a periodic signal based on the analysis result, And a subtracter configured to subtract the periodic signal from the signal output from the LTP analyzer, an abrupt noise reducer configured to attenuate the abrupt noise among the signals output from the LTP analyzer based on the abrupt noise signal output from the abrupt noise detector, An adder configured to add the signal output from the noise attenuator and the periodic signal output from the periodic signal generator; And A frequency converter for dividing the input digital signals into m blocks, converting the digital signals into the frequency domain in units of the divided blocks, and outputting m frequency domain signals converted into the frequency domain, A first energy calculator for calculating m energy values existing in a high frequency band for each of the m frequency domain signals input from the first energy calculator and outputting m first energy values, A first background signal energy calculator for estimating m first background signal energy values using first energy values and a first background signal energy calculator for calculating m first energy values and first background signal energy M pieces of the first background signal energy values are matched with the first energy values, A first comparator for comparing the first background signal energy values to detect first burst noise and outputting a first burst noise signal including the first burst noise; A second energy calculator for calculating a second energy value as a reference and outputting the second energy value; a second energy calculator for estimating second background signal energy values using the second energy values input from the second energy calculator; A first background energy calculator for calculating a first background energy value and a second background energy calculator for outputting second estimated background energy values, 2 background signal energy values, comparing the second energy values and the second background signal energy values to detect second burst noise, A second comparator for outputting a second burst noise signal including second burst noise and a second comparator for dividing the detected burst noise signal into m blocks and outputting the second burst noise signal And an unsteady noise detector for calculating an unexpected noise only when the number is greater than or equal to a predetermined number and outputting a third unexpected noise signal reflecting the calculated result, . 17. The method of claim 16, Wherein the sudden noise detector comprises: And a third comparator for comparing the third incident noise signal with the first incident noise signal inputted from the first comparator and then comparing the third incident noise signal and the third incident noise signal, And outputs a final sudden noise signal reflecting the calculated result. 17. The method of claim 16, Further comprising a controller for driving only the sudden noise eliminator without driving the sudden noise detector when a sudden noise generating signal for generating a sudden noise is inputted.

Images