JP6553561B2 - Signal analysis apparatus, method, and program - Google Patents

Description

  The present invention relates to a signal analysis apparatus, method, and program, and more particularly, to a signal analysis apparatus, method, and program for estimating parameters.

  The present invention addresses the problem of suppressing noise from speech signals. A technique for suppressing noise mixed in a voice signal is called a voice enhancement technique. Speech enhancement techniques are used in high-quality speech communication and preprocessing of speech recognition.

  There are three main approaches to speech enhancement. The unsupervised approach is the situation where both the target speech and the noise can not be informed in advance about its acoustic characteristics, while the semi-supervised approach is that the other speech of the same speaker is obtained in advance for the target speech. The supervised approach is a speech enhancement method that assumes a situation in which noise having the same characteristics as the target noise can be obtained in advance in addition to speech. The problem to be solved by the present invention is like a semi-supervised approach. As one of semi-supervised approaches, a method based on Semi-supervised Non-negative Matrix Factorization (SSNMF) has been proposed (Non-patent Document 1). In this method, the power spectrum of speech and noise can be estimated by fitting the observed spectrum at each time with a non-negative combination of the speech base spectrum and the noise base spectrum that have been learned in advance.

P. Smaragdis, B. Raj, and M. Shashanka, “Supervised and semi-supervised separation of sounds from single-channel mixtures,” in Proc. Independent Component Analysis and Signal Separation, pp. 414-421, 2007.

  While the technique of Non-Patent Document 1 can obtain enhanced speech with a high signal-to-noise ratio even in an unknown noise environment, it is not always necessary due to musical noise (noise components scattered in the time-frequency plane) consisting of residual noise components. There are many cases where the sound is not audibly good quality. In addition, since this method does not guarantee the natural temporal change of the speech spectrum, the spectrum of the enhanced speech tends to be discontinuous, which also causes a reduction in the auditory quality of the enhanced speech.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a signal analysis apparatus, method, and program capable of suppressing noise and emphasizing a high-quality audio signal.

  In order to achieve the above object, a signal analysis device according to the present invention receives observation time series data of an observation signal in which a speech signal and a noise signal are mixed and inputs an observation spectrogram representing an observation spectrum at each time and each frequency. Based on the parameters of the time-frequency expansion unit to output, the observation spectrogram output by the time-frequency expansion unit, and the mixed Gaussian distribution model of the speech signal learned in advance, the observation spectrum of each time and each frequency, From the spectrum of each time and each frequency obtained from the vocal tract spectrum sequence and the sound source spectrum sequence of the audio signal, the base spectrum representing the power spectrum at each base and each frequency of the noise signal, and the activation parameter representing the power at each time Spectrum of each required time and each frequency A distance from the sum of the vocal tract spectrum of the speech signal and a distance between the vocal tract spectrum series of the speech signal and a maximum likelihood spectrum series in the mixed Gaussian distribution model corresponding to the vocal tract spectrum series; a sound source spectrum series of the speech signal; It is obtained from the distance from the maximum likelihood spectrum sequence in the mixed Gaussian distribution model corresponding to the spectrum sequence, the observed spectrum at each time and each frequency of the noise signal, the base spectrum of the noise signal, and the activation parameter. The base spectrum and the activation parameter of the noise signal, the vocal tract spectrum sequence of the speech signal, and the speech so as to reduce the criterion expressed using the distance from the spectrum of each time and each frequency Source spectral sequence of the signal, and observation of each time and each frequency of the noise signal A parameter estimation unit that estimates the spectrum, is configured to include a.

  A signal analysis method according to the present invention is a signal analysis method in a signal analysis apparatus including a time-frequency expansion unit and a parameter estimation unit, wherein the time-frequency expansion unit is an observation in which an audio signal and a noise signal are mixed. The observation spectrogram representing the observation spectrum of each time and each frequency is output with the time series data of the signal as input, and the parameter estimation unit outputs the observation spectrogram output by the time frequency expansion unit, and the speech learned in advance. Based on the parameters of the mixed Gaussian distribution model of the signal, the spectrum of each time and each frequency obtained from the observed spectrum of each time and each frequency, the vocal tract spectrum sequence and the sound source spectrum sequence of the speech signal, and the noise signal Base spectrum representing the power spectrum at each base and each frequency and at each time Distance from the sum of the spectrum of each time and each frequency obtained from the activation parameter representing the power of the power, the vocal tract spectrum sequence of the speech signal, and the maximum likelihood in the mixed Gaussian distribution model corresponding to the vocal tract spectrum sequence Distance with a spectral sequence, distance between a source spectral sequence of the audio signal and a maximum likelihood spectral sequence in the mixed Gaussian distribution model corresponding to the source spectral sequence, and observation spectrum of each time and each frequency of the noise signal And the basis spectrum of the noise signal and the distance from the spectrum of each time and each frequency obtained from the activation parameter, the basis spectrum of the noise signal and the Activation parameters and vocal tracts of the speech signal Estimating and Le series, the sound source spectral trajectories of the audio signal, and observed spectrum at each time and each frequency of the noise signal.

  Moreover, the program of this invention is a program for functioning a computer as each part which comprises said signal analysis apparatus.

  As described above, according to the signal analysis device, method, and program of the present invention, the observed spectrum, the spectrum obtained from the vocal tract spectrum sequence and the sound source spectrum sequence, the base spectrum of the noise signal, and the activation parameter Distance with the sum of the spectrum determined from the distance, distance between the vocal tract spectrum sequence and the maximum likelihood spectrum sequence in the mixed Gaussian distribution model corresponding to the vocal tract spectrum sequence, the sound source spectrum sequence and the mixed Gaussian corresponding to the sound source spectrum sequence The distance from the maximum likelihood spectrum sequence in the distribution model, and the distance between the observed spectrum of each time and each frequency of the noise signal, and the spectrum of each time and each frequency determined from the base spectrum of the noise signal and the activation parameter Reduce the criteria expressed using Thus, the noise is suppressed by estimating the base spectrum and the activation parameters of the noise signal, the vocal tract spectrum sequence, the sound source spectrum sequence, and the observed spectrum of the noise signal, thereby suppressing noise and providing a high quality voice signal. Can be emphasized.

It is a block diagram showing functional composition of a signal analysis device concerning an embodiment of the invention. It is a flowchart figure which shows the learning process routine in the signal analyzer which concerns on embodiment of this invention. It is a flowchart figure which shows the parameter estimation processing routine in the signal analyzer which concerns on embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<Summary of invention>
First, an outline of the present embodiment will be described. In speech synthesis, the purpose is to synthesize aurally high quality speech, and in particular, a method of generating an optimum speech spectrum sequence using the statistical distribution of time derivative of the spectrum which greatly affects not only the spectrum but also the perception. Is taken. Also in speech enhancement for speech communication, since it is important how high-quality speech can be audibly produced, speech synthesis models and methods may be useful. The present invention proposes a method using the trajectory hidden Markov model proposed as a parametric speech synthesis model for the speech enhancement method of the SSNMF method for the problem of semi-supervised speech enhancement in an unknown noise environment.

<Principle of this embodiment>
Next, the principle of the present embodiment will be described.

<Formulation of problem>
The amplitude spectrogram or power spectrogram of the observation signal (hereinafter, observation spectrogram) is represented by _{Yω, t} . Where ω and t are indexes of frequency and time. Assuming that the spectrum is additive, the speech spectrum X ^(s) _{ω, t} and the noise spectrum X ⁽ⁿ⁾ _{ω, t} at each time are represented as L _S basis spectra, respectively.

And L _N basis spectra

Non-negative coupling of

It can be expressed by

SSNMF method pre-learned from clean speech learning samples

To the observed spectrum Y _{ω, t}

Is a method of estimating speech components and noise components included in an observation spectrogram by fitting (Non-patent Document 1). A speech signal can be obtained from the observation signal by a Wiener filter or the like from the speech spectrum and noise spectrum estimated values thus obtained. In this approach, the pre-learned speech base spectrum is a clue to separation of speech and noise, but the noise spectrum can be explained by the speech base spectrum and vice versa, so Y _{ω, t} and X _{Even if} the error of _{ω, t} can be reduced, X ^(S) _{ω, t} and X ^(N) _{ω, t} do not always correspond to the actual speech spectrum and noise spectrum. In addition, this system does not have a mechanism that can utilize statistics of time variation of speech spectrum in estimation of speech spectrum, and this is one factor that allows musical noise and discontinuous spectrum change. Conceivable. Therefore, in order to realize higher-quality speech enhancement, consider the tendency of temporal change of speech spectrum, give the same X _{ω, t,} and give the uncertainty of X ^(S) _{ω, t} and X ^(N) _{ω, t} We need stronger constraints to eliminate.

  In speech perception, it is known that the logarithmic spectrum of speech and its time derivative have a significant effect, and in speech synthesis, it is not only the logarithmic spectrum but also the logarithmic spectrum for the purpose of synthesizing aurally high quality speech. Is used to generate an optimal speech spectrum sequence using the statistical distribution of the time derivative (see Non-Patent Documents 2 and 3).

[Non-patent document 2]: T. Yoshimura, K. Tokuda, T. Kosashi, T. Kitamura, "Simultaneous modeling of spectrum, pitch and duration in HMM-based speech synthesis," in Proc. European Conference on Speech Communication and Technology (EUSIPCO 1999), vol. 5. pp. 2347-2350, 1999.

[Non-patent document 3]: H. Zen, K. Tokuda, T. Kitamura, "Reforming the HMM as a trajectory model by imposing explicit relationships between static and dynamic feature vector sequences," Computer Speech and Language, vol. 21, pp. 153-173, 2007.

This method is called Hidden Markov Model (HMM) speech synthesis. Even in speech enhancement for speech communication, it is important how a high quality speech can be audibly sensed, so models and methods of speech synthesis may be useful. Therefore, in the present invention, the logarithmic spectrum of speech is decomposed into vocal tract spectrum and vocal cord vibration spectrum components, and the regularization term for X ^(S) _{ω, t is} calculated based on the probability distribution of each component and the time derivative of each component. Design and give the regularization term by the conventional NMF for the logarithm spectrum of noise, and the sum of these regularization terms and the error criterion of Y ^(S) _{ω, t} and X ^(S) _{ω, t} We propose a parameter optimization algorithm as an optimization criterion.

The errors of Y _{ω, t} and X _{ω, t} are often measured by the square error, I divergence, Itakura-Saito-together pseudo-distance in SSNMF method, but in the present invention, the logarithmic spectral distance

Use Here, ω and t represent indexes of frequency and time. Next, based on the source filter theory of speech _, the spectrum X ^(S) _{ω, t} of the speech signal is

It is expressed as a product of the vocal tract spectrum F _{ω, t} and the source spectrum E _{ω, t} as in And for each of these

Consider the following criteria. here,

and

Represents the state sequences of different HMMs corresponding to the vocal tract spectrum sequence and the sound source spectrum sequence, respectively

and

Is the maximum likelihood spectrum sequence given the state sequence

Corresponds to This is the same form as the parameter generation method in HMM speech synthesis (see Non-Patent Documents 2 and 3 above).

Given in. However,

Is the state series

The average sequence of the state output distribution of each HMM, given

Is the state series

Is a diagonal matrix in which the variances of the state output distribution of each HMM given. Also,

Is a matrix that converts a vector in which a time series of parameters is stored into a vector obtained by combining itself and its time-value differential series. Also,

Represents a normal distribution with mean μ and variance covariance matrix Σ. Therefore, it can be seen that equations (7) and (8) mean the maximum likelihood spectrum series based on both probability distributions of a spectrum series determined by a given state series and state output distribution and its time derivative series. Here, the average and variance of the output distribution of each state of each HMM are constants previously learned from the vocal tract spectrum sequence and the sound source spectrum sequence of clean speech.

In the embodiment of the present invention, since a mixed Gaussian distribution model is used as a special case of HMM, the state sequence is

Represents a sequence of one state of different mixed Gaussian distribution models corresponding to the vocal tract spectrum sequence and the sound source spectrum sequence,

Is the mean sequence of the output distribution of each mixed Gaussian distribution model,

Is a diagonal matrix in which the variance of the output distribution of each mixed Gaussian distribution model is arranged in a diagonal component.

Furthermore, for X ^(N) _{ω, t}

Consider a criterion like In the proposed method, the weighted sum of the four criteria (3), (5), (6), and (11) is used.

The goal is to minimize However, α ₁ , α ₂ and α ₃ are weighting coefficients.

  As described above, how well this optimization problem can explain the speech spectrum sequence embedded in the observation spectrum sequence by the model (HQs (9) and (10)) that is isomorphic to the HMM speech synthesis generation model It is a problem to estimate while making a clue.

<Parameter estimation algorithm>

Although it is impossible to analytically obtain F, E, H ^(N) and U ^(N) which minimize ^N , it is possible to derive an iterative algorithm to search for the local optimal solution of the optimization problem based on the auxiliary function method .

In the optimization algorithm of the objective function F (θ) minimization problem by the auxiliary function method, first introduce the auxiliary variable 、,

Auxiliary functions satisfying

To design. If such an auxiliary function can be designed,

By alternately repeating the above, it is possible to obtain θ which locally minimizes the objective function F (θ). Below,

Derive the auxiliary function of and the update expression based on it.

about,

The auxiliary function is designed for each term on the right side of Equation (14). First, with regard to the first term, it is used that the following inequality holds.

  At x> 0, ξ> 0

And the equal sign holds only when x = ξ.

From the above inequality, for any _{ω ω, t} > 0

However, using the fact that the inverse function is a convex function, Jensen's inequality is

Holds. Where λ ^(S) _{ω, t} and λ ^(N) _{ω, t} are

Is an auxiliary variable that satisfies Summarizing these,

Get In addition, the condition for the equality of expression (21) is

It is.

Next, the auxiliary function of the second term of equation (14) is designed. Logy _{omega, t} so can take positive and negative _sign, logY ω, logX ω, make another inequality to _t depending on the sign of the _t. The negative logarithmic function is a convex function so Jensen's inequality

Get Where θ ^(S) _{ω, t} and θ ^(N) _{ω, t} are

Is a non-negative variable that satisfies Also, since the positive logarithmic function is a concave function

Holds. However, φ _{ω, t} is an arbitrary real number. These inequalities are

The equal sign is established when To summarize the above,

I can say. However, δ _x is an instruction function that is 1 when the condition x is satisfied, and 0 when the condition x is not satisfied.

next,

Design auxiliary functions for. First,

For the first term on the right side of Equation (31), using Equations (15), (16), and (17), for any η _{ω, t} > 0

And the equal sign is

It holds true for From the above,

Holds. Similarly,

Auxiliary function of with any _{ω ω, t} > 0

Can be designed by The equal sign is

It holds true for

Finally,

about

The first term on the right side of equation (37) can be written using equations (15), (16), and (17) for any ν _{ω, t} > 0

And the equal sign is

This holds true. Next, the second term auxiliary function is designed. logX _{^(N) ω,} because _t can take positive and negative _{^{sign, logX (N) ω, logΣ}} l H (N) ω in accordance with the sign of the _{^{t, t U (N) ω}} , different to _t Build an inequality. The negative logarithmic function is a convex function so Jensen's inequality

Get Where ψ _{ω, l, t} is

Is a non-negative variable that satisfies Also, since the positive logarithmic function is a concave function

Holds. However, π _{ω, t} is an arbitrary real number. These inequalities are

The equal sign is established when

For the third term, using equations (15), (16), and (17), for any ε _{ω, t} > 0

In addition, by Jensen's inequality

It becomes. Where σ _{ω, t} is any real variable, and τ _{ω, l, t} is

Is any non-negative variable that satisfies Summarizing these,

Get In addition, the equal sign of Formula (51) is

It holds true for

  To summarize the above,

And the right side

It can be an auxiliary function of By adding the above auxiliary functions

The auxiliary function is obtained. The update formula for each parameter that minimizes this auxiliary function is

It is obtained in closed form like. However,

It is.

<Configuration of Signal Analysis Device According to Embodiment of the Present Invention>
Next, the configuration of the signal analysis device according to the embodiment of the present invention will be described. As shown in FIG. 1, a signal analysis apparatus 100 according to an embodiment of the present invention is a ROM storing a CPU, a RAM, a program for executing a learning process routine and a parameter estimation process routine described later, and various data. And can be configured on a computer. Functionally, the signal analyzing apparatus 100 includes an input unit 10, an arithmetic unit 20, and an output unit 90 as shown in FIG.

  The input unit 10 receives time-series data of a clean sound signal (hereinafter referred to as a clean sound signal) that is not mixed with noise. The input unit 10 also receives time-series data of an acoustic signal (hereinafter referred to as an observation signal) in which an audio signal and a noise signal are mixed.

  The calculation unit 20 includes a time frequency expansion unit 24, a GMM parameter learning unit 32, a GMM parameter storage unit 34, a parameter estimation unit 36, and an audio signal generation unit 38.

  The time frequency expansion unit 24 calculates an amplitude spectrogram or a power spectrogram representing a spectrum of each frequency at each time based on the time series data of the clean speech signal. In the first embodiment, time-frequency expansion such as short-time Fourier transform and wavelet transform is performed.

In addition, the time-frequency expansion unit 24 calculates an observation spectrogram Y that is an amplitude spectrogram or a power spectrogram representing the observation spectrum Y _{ω, t} of each frequency ω at each time t based on the time-series data of the observation signal.

The GMM parameter learning unit 32 calculates the logarithm of the vocal tract spectrum and the vocal cord vibration spectrum of the speech based on the spectrum of each frequency at each time of the clean speech signal calculated by the time frequency expansion unit 24. Learn the parameters of the mixture Gaussian distribution model of the first-order difference amount. Specifically, the mean sequence of the output distribution of the mixed Gaussian distribution model

, A diagonal matrix in which the variances of the output distribution of the mixed Gaussian distribution model are arranged in diagonal components

To learn

The GMM parameter storage unit 34 is an average sequence of the output distribution of the mixed Gaussian distribution learned by the GMM parameter learning unit 32.

, A diagonal matrix in which the variances of the output distribution of the mixed Gaussian distribution model are arranged in diagonal components

Is remembered.

The parameter estimation unit 36 generates an observation spectrum Y of each time and each frequency based on the observation spectrogram Y output from the time frequency expansion unit 24 and the parameters of the mixed Gaussian distribution model stored in the GMM parameter storage unit 34. , Spectrum X ^(S) of each time and each frequency determined from vocal tract spectrum series F and sound source spectrum series E of speech signal, and each time and each frequency determined from the base spectrum of the noise signal and the activation parameter Distance with the sum X of the spectrum X ^(N) of

, The distance between the vocal tract spectral sequence F of the speech signal and the maximum likelihood spectral sequence given a sequence obtained by arranging one state of the mixed Gaussian distribution model corresponding to the vocal tract spectral sequence

, The distance between the source spectrum sequence E of the speech signal and the maximum likelihood spectrum sequence given a sequence obtained by arranging one state of the mixed Gaussian distribution model corresponding to the source spectrum sequence E

, And the distance between each spectrum of each time and each frequency obtained from the observed spectrum X ^(N) of each time and each frequency of the noise signal, and the base spectrum H ^{(N) of} the noise signal and the activation parameter U ^(N)

, The basis spectrum H ^(N) of the noise signal and the activation parameter U ^(N), and the vocal tract spectrum series F of the speech signal so as to reduce the criteria shown in the above equation (12) expressed using The source spectrum sequence E of the speech signal and the observed spectrum X ^(N) of each time and each frequency of the noise signal are estimated.

  Specifically, the parameter estimation unit 36 includes an initial value setting unit 40, an auxiliary variable update unit 42, a parameter update unit 44, and a convergence determination unit 46.

The initial value setting unit 40 includes an activation parameter U ^(S) of the audio signal, a base spectrum H ^(N) and an activation parameter U ^(N) of the noise signal, a vocal tract spectrum series F of the audio signal, and an audio signal. The initial value is set to the sound source spectrum sequence E and the observation spectrum X ^(N) of each time of each noise signal and each frequency. In addition, the state series

And sets a sequence in which one state is arranged, and based on the parameters of the mixed Gaussian distribution model stored in the GMM parameter storage unit 34, the maximum likelihood spectrum sequence

Set.

Auxiliary variable updating unit 42 is the initial value, or the last updated, and the vocal tract spectral sequence F of the audio signal, the sound source spectrum sequence E of the audio signal at each time and each frequency of the noise signal observed spectrum X ^{(N )} and on the basis, the (22) equation (23), (27), (28), (33), (36) where (39) equation (43) below, (44) (49) and (50) _, 、 _{ω, t} , λ ^(S) _{ω, t} , λ ^(N) _{ω, t} , θ ^(S) _{ω, t} for each frequency ω and each time t θ ^(N) _{ω, t} , φ _{ω, t} , η _{ω, t} , ζ _{ω, t} , ν _{ω, t} , π _{ω, t} , σ _{ω, t} , each base l, each frequency ω, and each time t Ψ _{l, ω, t} , τ _{ω, l, t} , each base l of noise signal and base spectrum H ^(N) _{ω, l of} each frequency ω, activation parameter U ^{(N )} Update _{l and t} .

The parameter updating unit 44 selects the observed spectrogram Y output from the time-frequency expanding unit 24, the respective frequencies ω updated by the auxiliary variable updating unit 42, and _{ω ω, t} , λ ^(S) _{ω, t with} respect to each time _t . λ ^(N) _{ω, t} , θ ^(S) _{ω, t} , θ ^(N) _{ω, t} , φ _{ω, t} , η _{ω, t} , ζ _{ω, t} , ν _{ω, t} , π _{ω, t} , σ _{ω, t} , each base l, each frequency ω, and ψ _{l, ω, t} , τ _{ω, l, t} with respect to each time t, vocal tract spectral series F of the speech signal which is an initial value or is previously updated Source spectrum sequence E of speech signal, observed spectrum X ^(N) of each time and frequency of noise signal, base spectrum H ^{(N) of} noise signal and activation parameter U ^(N) , state sequence

, Maximum likelihood spectral series

And the basis spectrum H ^(N) _{ω, l} of each base l and each frequency ω of the noise signal and the activation parameter U of each time t according to the equations (52) to (56) above. ^(N) Estimate _{l, t} , vocal tract spectral sequence F of speech signal, source spectrum sequence E of speech signal, and observed spectrum X ^(N) _{ω, t} of each time t of each noise signal and each frequency ω .

  The convergence determination unit 46 determines whether or not the convergence condition is satisfied, and repeats the update process in the auxiliary variable update unit 42 and the update process in the parameter update unit 44 until the convergence condition is satisfied.

  As the convergence condition, for example, the fact that the number of repetitions has reached the upper limit number can be used. Alternatively, as the convergence condition, it can be used that the difference between the value of the criterion of the above formula (12) and the value of the previous criterion is equal to or less than a predetermined threshold value.

The speech signal generation unit 38 generates a vocal tract spectrum sequence F of the speech signal acquired by the parameter estimation unit 36, a sound source spectrum sequence E of the speech signal, an observation spectrum X ^(N) of each time t of the noise signal and each frequency ω. _{Based on ω, t} and the observation spectrogram Y, an audio signal is generated according to the Wiener filter, and output from the output unit 90. For example, the observation spectrum X ^(S) _ω, of each time t of the audio signal and each frequency ω obtained from the vocal tract spectral sequence F of the audio signal acquired in the parameter estimation unit 36 and the sound source spectral sequence E of the audio signal _{. From t} , each time t of the noise signal acquired in the parameter estimation unit 36, the observed spectrum X ^(N) _{ω, t of} each frequency _ω, and the observed spectrogram Y, a voice signal is generated by the Wiener filter.

<Operation of Signal Analysis Device According to Embodiment of the Present Invention>
Next, the operation of the signal analysis device 100 according to the embodiment of the present invention will be described. First, when the time series data of the clean speech signal is received at the input unit 10, the signal analyzing apparatus 100 executes a learning processing routine shown in FIG.

  First, in step S100, the spectrum of each frequency at each time of the clean sound signal is calculated based on the time series data of the clean sound signal received by the input unit 10.

In step S106, based on the spectrum of each frequency at each time of the clean speech signal acquired in step S100, a mixed Gaussian distribution of the logarithm and logarithm of the primary difference amount is obtained for the vocal tract spectrum and vocal cord vibration spectrum of the speech. Learn model parameters. Specifically, the average series of the output distribution of the mixed Gaussian distribution model

, A diagonal matrix in which the variances of the output distribution of the mixed Gaussian distribution model are arranged in diagonal components

Is stored in the GMM parameter storage unit 34, and the learning process routine is terminated.

  Next, when the input unit 10 receives time-series data of an observation signal in which an audio signal and a noise signal are mixed, the signal analysis apparatus 100 executes a parameter estimation processing routine shown in FIG.

  First, in step S120, the observation spectrogram Y is calculated based on the time-series data of the observation signal received by the input unit 10.

In step S122, the activation parameter U ^{(S) of} the speech signal, the base spectrum H ^(N) and the activation parameter U ^(N) of the noise signal, the vocal tract spectrum sequence F of the speech signal, and the sound source spectrum of the speech signal Initial values are set to the sequence E and the observation spectrum X ^(N) of each time and each frequency of the noise signal.

In step S124, the vocal tract spectrum series F of the speech signal, the sound source spectrum series E of the speech signal, and the observation spectrum X of each time and each frequency of the noise signal, which are initial values or updated in step S125 described later. Based on ^(N) , (22), (23), (27), (28), (33), (36), (39), (43), (43), 44), (49), and (50), ζ _{ω, t} , λ ^(S) _{ω, t} , λ ^(N) _{ω, t} , θ ^(S) _ω, for each frequency ω and each time t _t , θ ^(N) _{ω, t} , φ _{ω, t} , _{ω ω, t} , _{ω ω, t} , _{ω ω, t} , π _{ω, t} , σ _{ω, t} , each base l, each frequency ω, and each Ψ _{l, ω, t} , τ _{ω, l,} t with respect to time t, the base spectrum H ^(N) _{ω, l} of each basis l of the noise signal and each frequency ω and the activation parameter U of each basis l and each time t ^(N) Update _{l, t} .

Next, in step S125, the observation spectrogram Y obtained in step S120 and ζ _{ω, t} , λ ^(S) _{ω, t} , λ ^{(N (N )} for each frequency ω and each time t updated in step S124. ⁾ _{ω, t} , θ ^(S) _{ω, t} , θ ^(N) _{ω, t} , φ _{ω, t} , η _{ω, t} , ζ _{ω, t} , ν _{ω, t} , π _{ω, t} , σ _{ω, t} , Each base l, each frequency ω, and ψ _{l, ω, t} , τ _{ω, l, t} with respect to each time t, the vocal tract spectrum series F of the speech signal which is an initial value or has been previously updated, and the speech Source spectrum sequence E of signal, observed spectrum X ^(N) of each time and frequency of noise signal, base spectrum H ^{(N) of} noise signal and activation parameter U ^(N) , state sequence

, Maximum likelihood spectral series

And the basis spectrum H ^(N) _{ω, l} of each base l and each frequency ω of the noise signal and the activation parameter U of each time t according to the equations (52) to (56) above. ^(N) Estimate _{l, t} , vocal tract spectral sequence F of speech signal, source spectrum sequence E of speech signal, and observed spectrum X ^(N) _{ω, t} of each time t of each noise signal and each frequency ω .

  Next, in step S128, it is determined whether the convergence condition is satisfied. If the convergence condition is satisfied, the process proceeds to step S130. If the convergence condition is not satisfied, the process proceeds to step S124, and the processes in steps S124 to S125 are repeated.

In step S130, the vocal tract spectrum series F of the speech signal finally updated in step S125, the sound source spectrum series E of the speech signal, and the observed spectrum X ^(N) of each time t of each noise signal and each frequency ω. _{Based on ω, t} and the observation spectrogram Y, an audio signal is generated according to the Wiener filter, output from the output unit 90, and the parameter estimation processing routine is ended.

<Experimental example>
Conducted an evaluation experiment to verify the noise suppression effect by the above-mentioned method using the voice data of NTT-AT multilingual voice database 2002 and RWCP noise data (four types of white noise, museum noise, babble noise, background music noise) It was. The comparison target was the conventional SSNMF method, and the improvement values of the signal-to-noise ratio (SNR) and mel frequency cepstrum coefficient distance before and after processing were evaluated. Test data was created by superimposing various noises on clean speech with various SNRs. All test data were monaural signals with a sampling frequency of 16 kHz, Fourier transform was performed for a short time with a frame length of 32 ms and a frame shift of 16 ms, and an observation spectrogram Y _{ω, t} was calculated. In learning, H ^(S) _{ω, l} was learned using a total of 500 sentences of 10 speakers (including 5 women and 5 men), and the vocal tract spectrum and vocal cord vibration spectrum The GMM parameters of the logarithm and the logarithmic primary difference amount were learned. The mixing number of GMM was 32. In the test, H ^(S) _{ω, l} and GMM parameters obtained by learning are fixed, and α ₁ = 1, α ₂ = 10, α ₃ = 1 and F _{ω, t} , E _{ω, t} , H ^{( N)} Estimated _{ω, l} , U ^(N) _{l, t} . After estimation, the estimated value of the speech signal was calculated by the Wiener filter using X ^(S) _{ω, t} and X ^(N) _{ω, t} . The initial values of the proposed algorithm are obtained by the conventional SSNMF.

  When the improved values of cepstrum distortion and SNR obtained by the proposed method and the conventional method under the above conditions were compared, the improved value of the proposed method was higher in most cases on any evaluation scale. That was confirmed.

  As described above, according to the signal analysis apparatus according to the embodiment of the present invention, the spectrum is obtained from the observed spectrum, the spectrum obtained from the vocal tract spectrum sequence and the sound source spectrum sequence, and the basis spectrum and activation parameters of the noise signal. Distance with the sum of the spectrums, the distance between the vocal tract spectrum sequence and the maximum likelihood spectrum sequence in the GMM corresponding to the vocal tract spectrum sequence, the sound source spectrum sequence, and the maximum likelihood spectrum sequence in the GMM corresponding to the sound source spectrum sequence And the criteria expressed using the distance between the observed spectrum of each time and frequency of the noise signal and the spectrum of each time and frequency obtained from the base spectrum and activation parameters of the noise signal So that the base spectrum and activation of the noise signal And ® emission parameters, and the vocal tract spectral series, and the sound source spectral series, by estimating the observed spectrum of the noise signal, it is possible to suppress the noise, to emphasize the high-quality voice signal.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made without departing from the scope of the present invention.

  For example, the processing for learning the GMM parameter and the parameter estimation for estimating the speech signal from the observation signal may be performed by different devices.

  In addition, since the order of the parameters to be updated is arbitrary, the order of the above embodiments is not limited.

  Further, in the present specification, the embodiment has been described in which the program is installed in advance. However, the program can be provided by being stored in a computer-readable recording medium or provided via a network. It is also possible.

DESCRIPTION OF SYMBOLS 10 Input part 20 Operation part 24 Time frequency expansion part 32 GMM parameter learning part 34 GMM parameter memory | storage part 36 Parameter estimation part 38 Speech signal generation part 40 Initial value setting part 42 Auxiliary variable update part 44 Parameter update part 46 Convergence determination part 90 Output Unit 100 Signal analyzer

Claims (7)

A time-frequency expansion unit for outputting an observation spectrogram representing an observation spectrum at each time and each frequency, using time-series data of the observation signal mixed with a speech signal and a noise signal as input,
Based on the observation spectrogram output by the time-frequency expansion unit and the parameters of the mixed Gaussian distribution model of the speech signal learned in advance,
Observation spectrum at each time and frequency, spectrum at each time and frequency obtained from the vocal tract spectrum sequence and sound source spectrum sequence of the audio signal, and base spectrum representing the power spectrum at each base and each frequency of the noise signal And the distance from the sum of the spectrum of each time and each frequency obtained from the activation parameter representing the power at each time,
A distance between a vocal tract spectrum sequence of the speech signal and a maximum likelihood spectrum sequence in the mixed Gaussian distribution model corresponding to the vocal tract spectrum sequence;
The distance between the sound source spectrum series of the voice signal and the maximum likelihood spectrum series in the mixed Gaussian distribution model corresponding to the sound source spectrum series, the observation spectrum of each time and each frequency of the noise signal, and the noise signal The distance from the spectrum of each time and each frequency obtained from the base spectrum and the activation parameter,
The basis spectrum and the activation parameter of the noise signal, the vocal tract spectrum sequence of the speech signal, the sound source spectrum sequence of the speech signal, and the noise signal A parameter estimator for estimating the observed spectrum at each time and each frequency;
Signal analyzer including: The signal analysis apparatus according to claim 1, wherein the criterion is expressed by the following equation.

Here, α ₁ , α ₂ and α ₃ represent predetermined weighting factors,

Represents the distance between the observed spectrogram Y and the sum X,

Is the distance between the vocal tract spectrum sequence F of the speech signal and the maximum likelihood spectrum sequence given the sequence s consisting of one state of the mixed Gaussian distribution model corresponding to the vocal tract spectrum sequence. Represent,

Is the distance between the sound source spectrum series E of the audio signal and the maximum likelihood spectrum series given a series q consisting of one state of the mixed Gaussian distribution model corresponding to the sound source spectrum series,

Is the spectrum of each time and frequency obtained from the observed spectrum X ^(N) of each time and frequency of the noise signal, the base spectrum H ^{(N) of the} noise signal and the activation parameter U ^(N). Represents the distance between The parameter estimation unit
The base spectrum and the activation parameter of the noise signal, the vocal tract spectrum sequence of the speech signal, the sound source spectrum sequence of the speech signal, and the auxiliary function that is an upper bound function of the criterion, A parameter updating unit that updates each time of the noise signal and the observed spectrum of each frequency;
A convergence determination unit that repeats the update by the parameter update unit until a predetermined convergence condition is satisfied;
The signal analysis device according to claim 1 or 2, comprising A signal analysis method in a signal analysis device including a time frequency expansion unit and a parameter estimation unit,
The time-frequency expansion unit inputs time-series data of an observation signal in which a speech signal and a noise signal are mixed, and outputs an observation spectrogram representing an observation spectrum at each time and each frequency,
The parameter estimation unit is based on the observation spectrogram output by the time-frequency expansion unit and a parameter of a mixed Gaussian distribution model of a speech signal learned in advance.
Observation spectrum at each time and frequency, spectrum at each time and frequency obtained from the vocal tract spectrum sequence and sound source spectrum sequence of the audio signal, and base spectrum representing the power spectrum at each base and each frequency of the noise signal And the distance from the sum of the spectrum of each time and each frequency obtained from the activation parameter representing the power at each time,
A distance between a vocal tract spectrum sequence of the speech signal and a maximum likelihood spectrum sequence in the mixed Gaussian distribution model corresponding to the vocal tract spectrum sequence;
The distance between the sound source spectrum series of the voice signal and the maximum likelihood spectrum series in the mixed Gaussian distribution model corresponding to the sound source spectrum series, the observation spectrum of each time and each frequency of the noise signal, and the noise signal The distance from the spectrum of each time and each frequency obtained from the base spectrum and the activation parameter,
The basis spectrum and the activation parameter of the noise signal, the vocal tract spectrum sequence of the speech signal, the sound source spectrum sequence of the speech signal, and the noise signal A signal analysis method that estimates the observed spectrum of each time and each frequency. The signal analysis method according to claim 4, wherein the criterion is expressed by the following equation.

Here, α ₁ , α ₂ and α ₃ represent predetermined weighting factors,

Represents the distance between the observed spectrogram Y and the sum X,

Is the distance between the vocal tract spectrum sequence F of the speech signal and the maximum likelihood spectrum sequence given the sequence s consisting of one state of the mixed Gaussian distribution model corresponding to the vocal tract spectrum sequence. Represent,

Is the distance between the sound source spectrum series E of the audio signal and the maximum likelihood spectrum series given a series q consisting of one state of the mixed Gaussian distribution model corresponding to the sound source spectrum series,

Is the spectrum of each time and frequency obtained from the observed spectrum X ^(N) of each time and frequency of the noise signal, the base spectrum H ^{(N) of the} noise signal and the activation parameter U ^(N). Represents the distance between In the estimation by the parameter estimation unit,
The parameter update unit reduces the base function of the noise signal and the activation parameter, the vocal tract spectrum sequence of the speech signal, and the sound source of the speech signal so that the auxiliary function that is an upper bound function of the criterion is reduced. Update the spectrum series and the observed spectrum at each time and each frequency of the noise signal,
The signal analysis method according to claim 4, wherein the convergence determination unit includes repeating the update by the parameter update unit until a predetermined convergence condition is satisfied.   The program for functioning a computer as each part of the signal-analysis apparatus in any one of Claims 1-3.