CN104936088A

CN104936088A - A Hybrid Virtual Bass Enhancement Processing Method

Info

Publication number: CN104936088A
Application number: CN201510192051.3A
Authority: CN
Inventors: 张斌; 方勇; 刘华平; 王红梅; 邹禾灿
Original assignee: University of Shanghai for Science and Technology
Current assignee: University of Shanghai for Science and Technology
Priority date: 2015-04-21
Filing date: 2015-04-21
Publication date: 2015-09-23

Abstract

The invention discloses a mixed virtual bass enhancing method. The method includes a signal processing process, and aims at enhancing the bass effect of audio signals. The method comprises the following steps that the same pathway signal of original audio is divided into two paths, high-pass filtering at the initial frequency of Fc is carried out on one path of the signals to obtain part of signals which is higher than the loudspeaker cutoff frequency Fc, time delay processing is carried out on the path of signals to obtain high-frequency signals, and low-pass filtering at the cutoff frequency Fc is carried out on the other path of signals to obtain low-frequency signals; the low-frequency signals are separated to obtain transient and stable components, and virtual bass processing are carried out on the transient and stable components respectively to obtain virtual low-frequency signals, and the virtual low-frequency signals are overlapped with high-frequency signals to obtain virtual bass enhanced audio signals. The audio signals after virtual bass processing reserve original quality with enhanced bass effect, and the audio loudness, fullness, deepness and spaciousness are enhanced.

Description

A Hybrid Virtual Bass Enhancement Processing Method

技术领域technical field

本发明涉及音频处理技术领域，具体涉及一种混合虚拟低音增强处理方法。The invention relates to the technical field of audio processing, in particular to a hybrid virtual bass enhancement processing method.

背景技术Background technique

随着多媒体设备的小型化、轻便化，嵌入在这些设备中的扬声器在尺寸上有很严格的限制，由于扬声器的体积限制，小型扬声器的低频重放能力很差，但是音频中的低频成分对听音感受起着很重要的作用，直接影响声音的洪亮度、丰满度、浑厚感和空间感。如何改善小型扬声器的低频表现是音频设计领域亟待解决的问题。With the miniaturization and portability of multimedia devices, the speakers embedded in these devices have strict restrictions on the size. Listening experience plays a very important role, directly affecting the loudness, fullness, thickness and sense of space of the sound. How to improve the low-frequency performance of small speakers is an urgent problem in the field of audio design.

传统的低频增强方法是采用音频均衡器直接增强低频能量，这种方法会导致扬声器效率降低、重放信号产生畸变，严重时甚至可能损坏扬声器系统。相比而言，虚拟低音增强技术是一种较为有效的解决方法。虚拟低音技术形成于心理声学理论，人感知重低音时，并不主要依靠低音的基频，而更多地是依靠该基频的各次谐波，即使低音信号的基频被抑制，只要其各次谐波以及这些谐波的关系依然存在，那么对于人耳来说，仍然能够感觉到低音效果。虚拟低音增强技术利用了这一现象，将低于扬声器截止频率的成分滤除，并适当增加其倍频点处的谐波能量，利用扬声器在谐波频段的重放能力较强的优势，可从主观听觉上虚拟出低音效果。The traditional method of low-frequency enhancement is to use an audio equalizer to directly enhance low-frequency energy. This method will reduce the efficiency of the speaker, distort the playback signal, and even damage the speaker system in severe cases. In comparison, virtual bass enhancement technology is a more effective solution. The virtual bass technology is formed from the theory of psychoacoustics. When people perceive heavy bass, they do not mainly rely on the fundamental frequency of the bass, but rely more on the harmonics of the fundamental frequency. Even if the fundamental frequency of the bass signal is suppressed, as long as its The harmonics and the relationship between these harmonics still exist, so for the human ear, the bass effect can still be felt. The virtual bass enhancement technology takes advantage of this phenomenon to filter out the components lower than the cut-off frequency of the speaker, and appropriately increase the harmonic energy at the octave point, taking advantage of the speaker's strong replay ability in the harmonic frequency band, it can Virtualize the bass effect from the subjective hearing.

现有的虚拟低音增强算法主要是基于非线性器件的和基于相位声码器的还有基于两者结合的混合虚拟低音增强算法，基于非线性器件的方法处理速度较快，对于瞬态信号所产生的低音效果较好，比如鼓声等，但是对于稳态信号会产生非线性畸变。基于相位声码器算法是通过短时傅理叶变换得到音频信号的时频信息，然后通过增加信号低频成分相位变化率的方式产生谐波。这种方法比较灵活，有效地控制了信号畸变，但是处理速度较慢，而且相位声码器为了得到更高的低频分辨率，必须在时域中有一个很大的分析窗口，这样会对瞬态信号产生一定的失真，所以基于相位声码器的低音增强算法处理稳态信号有较好的表现。混合的虚拟低音增强系统结合了非线性器件和相位声码器的优点，将音频信号分别通过两种谐波生成器，将得到的信号经过一个谐波能量控制模块以分配两者的比例。然而，由于输入到两种谐波生成器的音频信号是包含瞬态成分和稳态成分的原始信号，所以基于非线性器件生成的谐波会带有稳态信号的失真，而基于相位声码器生成的谐波会带有瞬态信号的失真，因此最后产生的虚拟低音信号包含了大量的失真，不能很好的还原真实的音质。The existing virtual bass enhancement algorithms are mainly based on nonlinear devices and phase vocoders, as well as hybrid virtual bass enhancement algorithms based on the combination of the two. The resulting bass effect is better, such as drums, etc., but it will produce nonlinear distortion for steady-state signals. Based on the phase vocoder algorithm, the time-frequency information of the audio signal is obtained through short-time Fourier transform, and then harmonics are generated by increasing the phase change rate of the low-frequency component of the signal. This method is more flexible and effectively controls the signal distortion, but the processing speed is slow, and the phase vocoder must have a large analysis window in the time domain in order to obtain higher low-frequency resolution, which will affect the transient Therefore, the bass enhancement algorithm based on the phase vocoder has better performance in dealing with steady-state signals. The hybrid virtual bass enhancement system combines the advantages of nonlinear devices and phase vocoders. The audio signal is passed through two kinds of harmonic generators respectively, and the obtained signal is passed through a harmonic energy control module to distribute the ratio of the two. However, since the audio signals input to the two kinds of harmonic generators are original signals containing transient components and steady-state components, the harmonics generated based on nonlinear devices will have distortion of steady-state signals, while the harmonics generated based on phase vocoding The harmonics generated by the amplifier will have the distortion of the transient signal, so the final virtual bass signal contains a lot of distortion, which cannot restore the real sound quality very well.

发明内容Contents of the invention

本发明的目的是为了解决现有技术存在的缺点和不足，提供一种新型的混合虚拟低音增强处理方法。该方法利用了声源分离算法，将原始音频信号中的瞬态和稳态信号先分离开后再分别进行相应的虚拟低音处理，克服了传统方法失真度高的不足，显著提高了虚拟低音增强的效果。The purpose of the present invention is to provide a novel hybrid virtual bass enhancement processing method in order to solve the shortcomings and deficiencies in the prior art. This method uses the sound source separation algorithm to separate the transient and steady-state signals in the original audio signal and then perform corresponding virtual bass processing, which overcomes the shortcomings of the traditional method of high distortion and significantly improves the virtual bass enhancement. Effect.

为了达到上述目的，本发明采用了下述技术方案：In order to achieve the above object, the present invention adopts following technical scheme:

一种混合虚拟低音增强处理方法，该信号处理方法首先将原始音频的同一通路信号分为两路，其中一路经过起始频率为Fc的高通滤波处理后得到高于扬声器截止频率Fc的部分，经延时处理后得到高频信号；另一路信号经过截止频率为Fc的低通滤波处理得到低频信号；将低频信号经过信号分离处理，得到瞬态成分和稳态成分；将瞬态成分和稳态成分分别经过相应的虚拟低音处理后得到虚拟低频信号，并与高频信号叠加，得到虚拟低音增强后的音频信号。其具体步骤如下：A hybrid virtual bass enhancement processing method, the signal processing method first divides the same path signal of the original audio frequency into two paths, wherein one path is processed by a high-pass filter with an initial frequency of Fc to obtain a part higher than the cut-off frequency Fc of the loudspeaker. The high-frequency signal is obtained after delay processing; the other signal is processed by a low-pass filter with a cutoff frequency of Fc to obtain a low-frequency signal; the low-frequency signal is subjected to signal separation processing to obtain a transient component and a steady-state component; the transient component and the steady-state The components are respectively processed by the corresponding virtual bass to obtain a virtual low-frequency signal, and superimposed with the high-frequency signal to obtain an audio signal after virtual bass enhancement. The specific steps are as follows:

(1)将原始音频的同一通路信号分为两路，其中一路经过高通滤波处理得到高于扬声器截止频率Fc的部分，再经延时处理后得到高频信号；另一路经过截止频率为Fc的低通滤波处理后得到低频信号。(1) The same channel signal of the original audio is divided into two channels, one of which is processed by high-pass filtering to obtain a part higher than the cut-off frequency Fc of the speaker, and then the high-frequency signal is obtained after delay processing; the other channel passes through the part with a cut-off frequency of Fc Low-frequency signals are obtained after low-pass filtering.

(2)将低频信号经过瞬态/稳态分离模块，得到低频信号的瞬态成分和稳态成分。(2) Pass the low-frequency signal through the transient/steady-state separation module to obtain the transient component and the steady-state component of the low-frequency signal.

(3)将低频信号的稳态成分经过基于相位声码器的虚拟低音生成模块，得到稳态成分的虚拟低音；将低频信号的瞬态成分经过基于非线性器件的虚拟低音生成模块，得到瞬态成分的虚拟低音。(3) pass the steady-state component of the low-frequency signal through the virtual bass generating module based on the phase vocoder to obtain the virtual bass of the steady-state component; pass the transient component of the low-frequency signal through the virtual bass generating module based on the nonlinear device to obtain the instantaneous virtual bass for dynamic components.

(4)根据等响度曲线对步骤(3)产生的稳态成分和瞬态成分的虚拟低音进行谐波幅度控制，使各谐波响度满足等响度曲线，并合称为一路，得到幅度控制后的虚拟低音信号。(4) Carry out harmonic amplitude control to the steady-state component and the virtual bass of the transient component produced in step (3) according to the equal-loudness curve, so that each harmonic loudness satisfies the equal-loudness curve, and collectively referred to as one path, after obtaining the amplitude control virtual bass signal.

(5)将步骤(1)得到的高频信号与步骤(4)得到的幅度控制后的虚拟低音信号合成为一路，得到虚拟低音处理后的音频信号。(5) Synthesizing the high-frequency signal obtained in step (1) with the amplitude-controlled virtual bass signal obtained in step (4) to obtain an audio signal after virtual bass processing.

本发明的一种混合虚拟低音增强处理方法与现有技术相比较，具有如下优势：Compared with the prior art, a kind of hybrid virtual bass enhancement processing method of the present invention has the following advantages:

本发明考虑了音频信号中普遍具有瞬态成分和稳态成分，首先利用瞬态/稳态信号分离算法将原始音频中的瞬态成分和稳态成分分离开来，再利用传统的基于非线性器件的虚拟低音生成模块和基于相位声码器的虚拟低音生成模块进行虚拟低音处理，大大减少了虚拟低音处理的稳态失真和瞬态失真，使虚拟低音处理后的音频信号在保留原始品质的同时增强了低音效果，增强了音频的洪亮度、丰满度、浑厚感和空间感。The present invention considers that audio signals generally have transient components and steady-state components, and first uses a transient/stationary signal separation algorithm to separate the transient components and steady-state components in the original audio, and then uses the traditional non-linear based The virtual bass generation module of the device and the virtual bass generation module based on the phase vocoder perform virtual bass processing, which greatly reduces the steady-state distortion and transient distortion of the virtual bass processing, so that the audio signal after the virtual bass processing retains the original quality. At the same time, the bass effect is enhanced, and the loudness, fullness, richness and sense of space of the audio are enhanced.

附图说明Description of drawings

图1为本发明混合虚拟低音增强处理方法的总流程框架示意图。Fig. 1 is a schematic diagram of the general flow framework of the hybrid virtual bass enhancement processing method of the present invention.

图2为图1中的虚拟低音处理模块的具体流程示意图。FIG. 2 is a schematic flow chart of the virtual bass processing module in FIG. 1 .

具体实施方式Detailed ways

下面结合实施例及附图对本发明作进一步的详细描述，但本发明的实施方式不限于此。The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

本发明的混合虚拟低音增强处理方法，音频信号中包含着瞬态成分和稳态成分，利用瞬态/稳态信号分离算法将原始音频信号中的瞬态成分和稳态成分分离开来，然后分别通过基于非线性器件的虚拟低音生成模块和基于相位声码器的虚拟低音生成模块进行虚拟低音处理，然后将得到的两路虚拟低音信号经过谐波幅度控制使其满足等响度曲线并合成为一路，最后将得到的虚拟低音信号与原始音频信号经高通滤波和延时处理后得到的高频信号进行叠加，得到混合虚拟低音增强处理后的音频信号。如图1、图2所示，具体实施步骤如下：In the mixed virtual bass enhancement processing method of the present invention, the audio signal contains transient components and steady components, and the transient components and steady components in the original audio signal are separated by using a transient/steady signal separation algorithm, and then The virtual bass processing is carried out through the virtual bass generation module based on the nonlinear device and the virtual bass generation module based on the phase vocoder, and then the obtained two-way virtual bass signals are controlled by harmonic amplitude to meet the equal loudness curve and synthesized into All the way, finally the obtained virtual bass signal is superimposed with the high-frequency signal obtained after high-pass filtering and delay processing of the original audio signal to obtain the audio signal after mixed virtual bass enhancement processing. As shown in Figure 1 and Figure 2, the specific implementation steps are as follows:

(1)将原始音频信号x_ori(t)分为两路，其中一路通过起始频率为Fc的高通滤波器，得到高于扬声器截止频率Fc的信号，然后经过延时处理模块得到原始信号的高频信号x_H(t)，另一路通过截止频率为Fc的低通滤波器，得到低于扬声器截止频率Fc的低频信号x_L(t)；(1) Divide the original audio signal x _ori (t) into two channels, one of which passes through a high-pass filter with a starting frequency of Fc to obtain a signal higher than the cut-off frequency Fc of the loudspeaker, and then obtains the original signal through a delay processing module The high-frequency signal x _H (t), the other way passes through a low-pass filter with a cut-off frequency of Fc to obtain a low-frequency signal x _L (t) lower than the cut-off frequency of the loudspeaker Fc;

(2)将低频信号x_L(t)输入到虚拟低音处理模块，得到虚拟低音信号x′_L(t)，具体实施步骤如下：(2) Input the low-frequency signal x _L (t) to the virtual bass processing module to obtain the virtual bass signal x′ _L (t), and the specific implementation steps are as follows:

(2-1)将低频信号x_L(t)经过HPSS声源分离算法处理后可将瞬态信号和稳态信号从原始信号中分离出来，得到稳态信号x_{L_H}(t)，和瞬态信号x_{L_P}(t)。具体分离算法如下：(2-1) After the low-frequency signal x _L (t) is processed by the HPSS sound source separation algorithm, the transient signal and the steady-state signal can be separated from the original signal to obtain the steady-state signal x _{L_H} (t), and the transient Signal x _{L_P} (t). The specific separation algorithm is as follows:

a.将低频信号x_L(t)进行短时傅里叶变换(STFT)，得到低频信号的时频谱矩阵，记为是一个N×K的复数矩阵，其中N表示时频谱总共的时间样点数，K表示每个时间样点所拥有的频率样点数。a. Perform short-time Fourier transform (STFT) on the low-frequency signal x _L (t) to obtain the time-spectrum matrix of the low-frequency signal, denoted as is an N×K complex number matrix, where N represents the total number of time samples in the time spectrum, and K represents the number of frequency samples in each time sample.

b.对原始信号时频谱进行取模运算，得到其幅度谱Y，Y是一个N×K的实数矩阵，Y_n,k表示时频谱幅度谱中的某一个时频点，n和k都是整数，代表着该时频点的时间索引和频率索引；b. Spectrum of the original signal Perform a modulo operation to obtain its amplitude spectrum Y, Y is a real number matrix of N×K, Y _n,k represents a certain time-frequency point in the time-frequency spectrum amplitude spectrum, n and k are both integers, representing the time-frequency The time index and frequency index of the point;

c.设置稳态信号幅度谱H和瞬态信号幅度谱P的初始值都为其中γ为衰减因子，在本实施例中γ＝0.5；c. Set the initial values of the steady-state signal amplitude spectrum H and the transient signal amplitude spectrum P to be Wherein γ is an attenuation factor, in this embodiment γ=0.5;

d.利用如下公式进行迭代处理：d. Use the following formula for iterative processing:

${H h}_{n no,, k k}^{γ γ} &LeftArrow; &LeftArrow; \frac{{H h}_{n no,, k k}^{((n no - - mean mean))}}{\sqrt{{(({H h}_{n no,, k k}^{((n no - - mean mean))}))}^{22} + + {(({P P}_{n no,, k k}^{((n no - - mean mean))}))}^{22}}} {Y Y}_{n no,, k k}^{γ γ} - - - - - - ((11))$

${P P}_{n no,, k k}^{γ γ} &LeftArrow; &LeftArrow; \frac{{P P}_{n no,, k k}^{((n no - - mean mean))}}{\sqrt{{(({H h}_{n no,, k k}^{((n no - - mean mean))}))}^{22} + + {(({P P}_{n no,, k k}^{((n no - - mean mean))}))}^{22}}} {Y Y}_{n no,, k k}^{γ γ} - - - - - - ((22))$

其中，in,

${H h}_{n no,, k k}^{((n no - - mean mean))} = = \frac{11}{22 {N N}^{' '}} {Σ Σ}_{{n no}^{' '} = = 11}^{{N N}^{' '}} (({H h}_{n no + + {n no}^{' '},, k k}^{γ γ} + + {H h}_{n no - - {n no}^{' '},, k k}^{γ γ})) - - - - - - ((33))$

${P P}_{n no,, k k}^{((n no - - mean mean))} = = \frac{11}{22 {K K}^{' '}} {Σ Σ}_{{k k}^{' '} = = 11}^{{K K}^{' '}} (({P P}_{n no,, k k + + {k k}^{' '}}^{γ γ} + + {P P}_{n no,, k k - - {k k}^{' '}}^{γ γ})) - - - - - - ((44))$

e.将得到的稳态信号幅度谱H和瞬态信号幅度谱P用如下公式得到维纳滤波器；E. the obtained steady-state signal amplitude spectrum H and the transient signal amplitude spectrum P obtain the Wiener filter with the following formula;

${W W}_{H h} = = \frac{H h}{\sqrt{{H h}^{22} + + {P P}^{22}}} - - - - - - ((55))$

${W W}_{P P} = = \frac{P P}{\sqrt{{H h}^{22} + + {P P}^{22}}} - - - - - - ((66))$

f.将原始信号时频谱乘以维纳滤波器，得到稳态信号时频谱和瞬态信号时频谱计算公式如下；f. The time spectrum of the original signal Multiply by the Wiener filter to get the frequency spectrum of the steady-state signal and the spectrum of the transient signal Calculated as follows;

$\overset{^^}{H h} = = \overset{^^}{Y Y} \times \times {W W}_{H h} - - - - - - ((77))$

$\overset{^^}{P P} = = \overset{^^}{Y Y} \times \times {W W}_{P P} - - - - - - ((88))$

g.对得到的稳态信号时频谱和瞬态信号时频谱进行逆短时傅里叶变换(iSTFT)，得到时域的稳态信号x_{L_H}(t)和时域的瞬态信号x_{L_P}(t)。g. The time spectrum of the obtained steady-state signal and the spectrum of the transient signal The inverse short-time Fourier transform (iSTFT) is performed to obtain the steady-state signal x _{L_H} (t) in the time domain and the transient signal x _{L_P} (t) in the time domain.

(2-2)将上述瞬态信号x_{L_P}(t)输入到基于非线性器件的虚拟低音生成模块，得到瞬态信号的谐波信号x′_{L_P}(t)。基于非线性器件的虚拟低音生成模块是利用乘法器元件来产生低频成分的高次谐波，若乘法器两端输入信号都是频率为f₀的纯音信号：(2-2) Input the above-mentioned transient signal x _{L_P} (t) to the virtual bass generating module based on the nonlinear device to obtain the harmonic signal x′ _{L_P} (t) of the transient signal. The virtual bass generation module based on nonlinear devices uses multiplier components to generate high-order harmonics of low-frequency components. If the input signals at both ends of the multiplier are pure tone signals with a frequency of f ₀ :

$X x = = {Ae Ae}^{- - j j 22 π π {f f}_{00} t t} - - - - - - ((99))$

则乘法器的输出信号是频率为2f₀的纯音信号：Then the output signal of the multiplier is a pure tone signal with frequency 2f ₀ :

$\begin{matrix} {X x}_{out out} = = X x \times \times X x \\ = = {Ae Ae}^{- - j j 22 π π {f f}_{00} t t} \times \times {Ae Ae}^{- - j j 22 π π {f f}_{00} t t} \\ = = {A A}^{22} {e e}^{- - 22 π π ((22 {f f}_{00})) t t} \end{matrix} - - - - - - ((1010))$

同理，若输入信号是频率分别为f₀与2f₀的纯音，则输出信号的频率为3f₀；Similarly, if the input signals are pure tones with frequencies f ₀ and 2f ₀ respectively, the frequency of the output signal is 3f ₀ ;

(2-3)将上述稳态信号x_{L_H}(t)输入到基于相位声码器的虚拟低音生成模块，得到稳态信号的谐波信号x′_{L_H}(t)。基于相位声码器的虚拟低音生成模块是利用相位声码器来控制谐波，相位声码器算法是基于短时傅里叶变换利用相位信息来实现信号的音调变换的方法，可以应用于音频的时间拉伸和压缩。基于相位声码器的虚拟低音生成模块首先对输入稳态信号x_{L_H}(t)进行短时傅里叶变换，短时傅里叶变换会通过加窗处理将信号分段，对每段信号进行FFT，得到的信号可由一组正弦信号组合而成，改变正弦信号的相位变化率，其频率就会发生改变，从而生成高次谐波，然后再对每段信号进行逆FFT操作，还原成时域信号，再将每段信号相加，得到稳态信号的谐波信号x′_{L_H}(t)；(2-3) Input the above-mentioned steady-state signal x _{L_H} (t) to the virtual bass generating module based on the phase vocoder to obtain the harmonic signal x′ _{L_H} (t) of the steady-state signal. The virtual bass generation module based on the phase vocoder uses the phase vocoder to control the harmonics. The phase vocoder algorithm is based on the short-time Fourier transform and uses the phase information to realize the pitch transformation of the signal, which can be applied to audio time stretching and compression. The virtual bass generation module based on the phase vocoder first performs short-time Fourier transform on the input steady-state signal x _{L_H} (t), and the short-time Fourier transform will segment the signal through windowing processing, and perform FFT, the obtained signal can be composed of a group of sinusoidal signals, changing the phase change rate of the sinusoidal signal, its frequency will change, thereby generating higher harmonics, and then perform an inverse FFT operation on each segment of the signal to restore it to time domain signal, and then add each segment of the signal to obtain the harmonic signal x′ _{L_H} (t) of the steady-state signal;

(2-4)将上述的瞬态谐波信号x′_{L_P}(t)和稳态谐波信号x′_{L_H}(t)进行谐波幅度控制并合成为一路，由于不同频率的纯音信号和具有不同频率组分的复音信号，即使声压级相同，响度可不相同，需要根据等响度曲线对生成的谐波成分进行谐波幅度控制，使其达到理想的低音效果，然后合成为一路，得到虚拟低音信号x′_L(t)。(2-4) The above-mentioned transient harmonic signal x′ _{L_P} (t) and steady-state harmonic signal x′ _{L_H} (t) are subjected to harmonic amplitude control and synthesized into one channel. Since pure tone signals of different frequencies have different For polyphonic signals of frequency components, even if the sound pressure level is the same, the loudness can be different. It is necessary to control the harmonic amplitude of the generated harmonic components according to the equal loudness curve to achieve the ideal bass effect, and then synthesize them into one channel to obtain a virtual bass Signal _x'L (t).

(3)将步骤(2)得到的虚拟低音信号x′_L(t)和步骤(1)得到的高频信号x_H(t)混合成一路信号，即可得到经混合虚拟低音处理后的信号x_Bass(t)。(3) Mix the virtual bass signal x′ _L (t) obtained in step (2) and the high-frequency signal x _H (t) obtained in step (1) into one signal, and the signal processed by the mixed virtual bass can be obtained x _Bass (t).

综上所述，本发明提供了一种混合虚拟低音增强处理方法，该方法首先利用声源分离算法，将原始音频信号分离成瞬态和稳态两部分，再针对每一部分信号的特征对其进行相应的虚拟低音生成处理，在保证原始音频音质的同时，有效降低了虚拟低音增强处理后的失真，增强了音频信号的低音效果。In summary, the present invention provides a hybrid virtual bass enhancement processing method, the method first uses the sound source separation algorithm to separate the original audio signal into two parts, transient and steady state, and then compares the characteristics of each part of the signal Corresponding virtual bass generation processing is performed to effectively reduce the distortion after the virtual bass enhancement processing while ensuring the original audio quality, and enhance the bass effect of the audio signal.

Claims

1. a hybrid virtual bass enhancement processing method, is characterized in that, the method comprises the following processing procedures:

1) Divide the original audio signal x _ori (t) into two channels, one of which passes through a high-pass filter with a starting frequency of Fc to obtain a part higher than the cut-off frequency Fc of the speaker, and obtains a high-frequency signal x _H after delay processing (t), another road signal obtains low-frequency signal x _L (t) through the low-pass filter processing that cut-off frequency is Fc;

2) Carry out signal separation processing to described low-frequency signal x _L (t), obtain the steady-state signal x _{L_H} (t) of low-frequency audio signal and transient signal x _{L_P} (t);

3) The steady-state signal x _{L_H} (t) and the transient signal x _{L_P} (t) of the low-frequency signal are processed to generate virtual bass harmonics to obtain the steady-state harmonic signal x′ _{L_H} (t) and the transient harmonic signal x′ _{L_P} (t);

4) The steady-state harmonic signal x′ _{L_H} (t) and the transient harmonic signal x′ _{L_P} (t) are subjected to harmonic amplitude control processing, and then synthesized into one path to obtain a virtual bass signal x′ _L (t) ;

5) Synthesizing the high-frequency signal x _H (t) and the virtual bass signal x′ _L (t) to obtain the audio signal x _Bass (t) after virtual bass enhancement.

2. a kind of hybrid virtual bass enhancement processing method according to claim 1, is characterized in that, described low-frequency signal x _L (t) in described step 2) is carried out signal separation process and comprises the following steps:

a) performing short-time Fourier transform on the low-frequency signal x _L (t), to obtain the time spectrum of the low-frequency signal

b) time spectrum for the low frequency signal Carry out the modulo operation to obtain the amplitude spectrum Y of the frequency spectrum of the low frequency signal;

c) Set the initial values of the steady-state signal amplitude spectrum H and the transient signal amplitude spectrum P to be where γ is the attenuation factor;

d) Iterative processing is performed using the following formula:

{H h}_{n no,, k k}^{γ γ} &LeftArrow; &LeftArrow; \frac{{H h}_{n no,, k k}^{((n no - - mean mean))}}{\sqrt{{(({H h}_{n no,, k k}^{((n no - - mean mean))}))}^{22} + + {(({P P}_{n no,, k k}^{((n no - - mean mean))}))}^{22}}} {Y Y}_{n no,, k k}^{γ γ} - - - - - - ((11))

{P P}_{n no,, k k}^{γ γ} &LeftArrow; &LeftArrow; \frac{{P P}_{n no,, k k}^{((n no - - mean mean))}}{\sqrt{{(({H h}_{n no,, k k}^{((n no - - mean mean))}))}^{22} + + {(({P P}_{n no,, k k}^{((n no - - mean mean))}))}^{22}}} {Y Y}_{n no,, k k}^{γ γ} - - - - - - ((22))

in,

{H h}_{n no,, k k}^{((n no - - mean mean))} = = \frac{11}{22 {N N}^{' '}} {Σ Σ}_{{n no}^{' '} = = 11}^{{N N}^{' '}} (({H h}_{n no + + {n no}^{' '}}^{γ γ} + + {H h}_{n no - - {n no}^{' '},, k k}^{γ γ})) - - - - - - ((33))

{P P}_{n no,, k k}^{((n no - - mean mean))} = = \frac{11}{22 {K K}^{' '}} {Σ Σ}_{{k k}^{' '} = = 11}^{{K K}^{' '}} (({P P}_{n no,, k k + + {k k}^{' '}}^{γ γ} + + {P P}_{{n no,, k k - - k k}^{' '}}^{γ γ})) - - - - - - ((44))

E) obtaining the steady-state signal amplitude spectrum H and the transient signal amplitude spectrum P with the following formula to obtain the Wiener filter;

{W W}_{H h} = = \frac{H h}{\sqrt{{H h}^{22} + + {P P}^{22}}} - - - - - - ((55))

{W W}_{P P} = = \frac{P P}{\sqrt{{H h}^{22} + + {P P}^{22}}} - - - - - - ((66))

f) time spectrum of the original signal Multiply by the Wiener filter to get the frequency spectrum of the steady-state signal and the spectrum of the transient signal Calculated as follows;

\overset{^^}{H h} = = \overset{^^}{Y Y} \times \times {W W}_{H h} - - - - - - ((77))

\overset{^^}{P P} = = \overset{^^}{Y Y} \times \times {W W}_{P P} - - - - - - ((88))

g) Time spectrum for the obtained steady-state signal and the spectrum of the transient signal The inverse short-time Fourier transform is performed to obtain the steady-state signal x _{L_H} (t) in the time domain and the transient signal x _{L_P} (t) in the time domain.

3. a kind of mixing virtual bass enhancement processing method according to claim 1, is characterized in that, described step 3) in the virtual bass harmonic generation processing process refers to described transient signal x _{L_P} (t) input to a harmonic generator consisting of a multiplier element and a feedback loop to obtain the harmonic signal x′ _{L_P} (t) of the transient signal; the steady-state signal x _{L_H} (t) is input to the phase vocoder The harmonic generator is used to obtain the harmonic signal x′ _{L_H} (t) of the steady-state signal.

4. a kind of hybrid virtual bass enhancement processing method according to claim 1, is characterized in that, the harmonic amplitude control processing process in described step 4) refers to, according to equal loudness curve, to described transient harmonic The signal x′ _{L_P} (t) and the steady-state harmonic signal x′ _{L_H} (t) are adjusted to achieve the ideal bass effect, and then synthesized into one channel to obtain the virtual bass signal x′ _L (t).