WO2014059890A1 - 在线迭代的声重放系统频响均衡方法和装置 - Google Patents
在线迭代的声重放系统频响均衡方法和装置 Download PDFInfo
- Publication number
- WO2014059890A1 WO2014059890A1 PCT/CN2013/084956 CN2013084956W WO2014059890A1 WO 2014059890 A1 WO2014059890 A1 WO 2014059890A1 CN 2013084956 W CN2013084956 W CN 2013084956W WO 2014059890 A1 WO2014059890 A1 WO 2014059890A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- equalizer
- signal
- parameter
- iterative
- sound source
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R3/00—Circuits for transducers, loudspeakers or microphones
- H04R3/02—Circuits for transducers, loudspeakers or microphones for preventing acoustic reaction, i.e. acoustic oscillatory feedback
Definitions
- the present invention relates to a frequency response equalization method and apparatus for an acoustic playback system, and more particularly to an online iterative audio reproduction system frequency response equalization method and apparatus.
- This technology uses a single microphone to collect the impulse response data of the speaker system point by point in front of the speaker unit according to the serpentine path.
- the frequency response equalization of the speaker system is implemented by a specific response data processing algorithm.
- the company's room equalization technology, Room Perfect uses a microphone to capture the response data of the speaker system at multiple points in the room, and the multi-point response information to complete the listener's position. The response curve is balanced.
- KRK Corporation of the United States also introduced Ergo (Enhanced Room Geometry Optimization), an indoor sound reproduction system, which uses a single microphone to collect response data around the listener's location and process the response data to obtain equalizer parameters. The response to the listener's location point is balanced.
- the above-mentioned plurality of sound reproduction system equalization products are used for equalizing the speaker unit in the free field or the speaker system in the equalization room.
- the impulse response sequence of the speaker unit has a short duration and a small pulse tailing.
- the design of the equalizer is relatively simple, but the impulse response sequence of the speaker unit in the room lasts for a long time, and the pulse tailing is very serious.
- the design of the equalizer is complicated, and it is necessary to combine the response data of multiple spatial position points for comprehensive processing.
- the equalization algorithm of these equalization products essentially analyzes the acquired time domain response sequence and decomposes it into the minimum phase response part and the all-pass response part, and directly responds to the minimum phase response part.
- the poles of the inverse filter response can be obtained to obtain the parameters of the inverse filter response.
- the phase response is constant in the desired frequency band region, it does not change with the frequency. Then the phase-constant all-pass response portion has no significant effect on the tone of the reproduced signal and can be ignored, but if the phase response of the all-pass response portion varies with frequency in the desired frequency band, then the phase response The all-pass response portion of the frequency change causes a significant change in the tone of the playback signal, which needs to be considered for equalization.
- the traditional acoustic reproduction system equalization method is based on the analysis of the system impulse response function, fitting the zero-pole model of these impulse response functions, and then finding the inverse filter response of the system through the pole-zero inversion, thus obtaining the sound.
- the equalizer parameters of the playback system The parameter estimation process of these methods relies on the Lean Mean Squares (LMS) algorithm or the linear predictive coding (LPC) algorithm to calculate the parameters of the equalizer, which is based on the single estimation method.
- LMS Lean Mean Squares
- LPC linear predictive coding
- the peak-valley fluctuation characteristics of these equalized frequency response curves are caused by the parameter error of the equalizer.
- certain error defects exist in system equalizer parameter estimation, and it is necessary to consider multiple iterative estimation methods to make the estimated cascade by multiple iteration operations.
- the equalizer response gradually approaches the ideal system inverse filter response, reducing the parameter estimation error of the equalizer, thereby ensuring that the equalized frequency response curve has better flatness characteristics.
- An online iterative sound reproduction system frequency response equalization method includes the following steps:
- the noise signal in step (1) may be a white noise sequence of a specified bandwidth or a noise signal generated by a maximum length sequence (MLS), and the signal exhibits a flat power spectrum characteristic, Used for yii to develop a transfer function from the speaker to the microphone position in the free field or reverberant field.
- MLS maximum length sequence
- step ( ⁇ ) the sensor collects and records the feedback signal, as shown in FIG. 2, the type of the sensor is determined according to the equalization target of the sound reproduction system; if the sound reproduction system only needs equalization signal processing And the power amplifier drives the frequency response characteristics of the circuit components formed by the two parts, and the sensor can be a cable or analog to digital converter to obtain an impulse response sequence signal of the circuit processing portion generated by the noise excitation; if the sound reproduction system The response fluctuation characteristic after the coupling of the equalization circuit portion and the speaker portion is required, and the sensor is a microphone placed at a desired position in the space for recording the sound field propagation from the speaker to the microphone generated by the click excitation.
- the impulse response sequence if the acoustic reproduction system requires the response fluctuation characteristic of the equalization circuit portion, the speaker portion and the external environment portion together, the sensor is a microphone placed at a certain point in the space, the microphone Need to change the location of the space, record multiple spatial location points in turn ⁇ noise excitations correspondence between the speaker to the microphone in the sound field from propagating impulse response sequence.
- the LMS criterion of the step-by-step iteration in step (2) sequentially estimates the parameters of the equalizers of each stage.
- the estimation process of the equalizer parameters of the LMS criterion based on the stepwise iteration is as follows:
- ⁇ is the number of samples of the time-domain discrete sequence of the noise source signal; assuming the time domain impulse response sequence of the acoustic playback system
- M is the sequence length of the system time domain impulse response
- the length of the time domain impulse response sequence of the first equalizer to be solved is assumed to be L (L> M), then the response sequence vector can be expressed as:
- the feedback signal is processed by the first equalizer / ⁇ to be obtained, and the feedback signal expression after the first equalization is obtained.
- the mean square error between the sound source signal s w and the equalized feedback signal is calculated to take the most ⁇ value
- the equalizer parameter vector of the time as the estimated value of the first optimal equalizer parameter vector designed at the first iteration, the expression is:
- the estimated value of the optimal equalizer parameter vector can be calculated according to the LMS algorithm.
- the calculation process of the optimal equalizer parameters is based on LMS.
- the iterative algorithm is implemented.
- the calculation expression of the first optimal equalizer parameter is as follows - r ⁇ '(n) ⁇ 2 ⁇ J (n) ' r("- ) (") Two (" A) ) (") i gi 0? + 1) - heqjin) + 2 ⁇ ' e ⁇ (n) . ri ? - /), 0 ⁇ ⁇ L - 1 '
- ⁇ represents the delay amount of the sound source signal Sw (n);
- tr (R) is the trace of the matrix R
- RE ( r . r , is the conjugate transpose vector of the vector "
- ⁇ " is a time-domain sequence vector of length L formed by the feedback signal received by the sensor
- the vector expression of the i-th sample time feedback signal r ⁇ is: r(i) - ⁇ r(i) r(i-1)... r(i -L + 1)]
- ⁇ Length ⁇ is selected according to the following relationship:
- the actual value of the step factor ⁇ should be selected in combination with the actual operation of the LMS algorithm to ensure that the algorithm has a good convergence speed under the premise of small offset error.
- the calculation of the optimal equalizer parameters can also use the normalized LMS algorithm, namely the NLMS (Normaiized Least Mean Squares) algorithm.
- the backbone NLM:S algorithm The calculation expression of the first optimal equalizer parameter is as follows: r eq '( ⁇ ) 2 ⁇ ( ⁇ ) ' r ( n .
- ⁇ is the step factor and ⁇ is the small normal number.
- the values of the parameters ⁇ and Y should also be selected in combination with the actual operation of the algorithm to ensure that the convergence speed of the algorithm is maximized in the case of small offset errors.
- " is the transposition vector of the vector £ ⁇ ; after the feedback signal is processed by the ⁇ th equalizer, the expression of the feedback signal after the first equalization is: '(.)... 1 )] After the equalization of the feedback signal) and the sound source signal s w , we can calculate the eq after processing by the level equalizer
- step a continue to complete the parameter estimation of the first, second, ..., k L equalizers, assuming that the k-th equalizer parameter estimation is completed, such as As shown in Figure 6, the estimation process of the kth equalizer parameters is as follows:
- the feedback signal after filtering by the first k-1 equalizer is expressed as: ⁇ ⁇ [ ⁇ i0) — "(i)... ⁇ — 1)]
- the expression of the feedback signal processed by the first k equalizers is:
- the equalization when the mean square error between the sound source signal s w and the equalized feedback signal is minimized is calculated.
- the parameter vector as the estimated value of the kth optimal equalizer parameter vector designed for the kth iteration, whose expression is -
- LMS algorithm can calculate the estimated value of the optimal equalizer parameter vector, and the calculation process of the optimal equalizer parameter is realized by the LMS-based iterative algorithm, in the first iteration process
- the calculation expression of the kth optimal equalizer parameter is as follows:
- step size ⁇ is selected according to the following relationship:
- a raax is the maximum eigenvalue of the matrix R; the actual value of the step factor ⁇ should be selected in combination with the actual operation of the LMS algorithm to ensure that the algorithm has better under the premise of less offset error. convergence speed.
- the calculation of the optimal equalizer parameters can also use the normalized LMS algorithm, that is, the NLMS algorithm.
- the k-th optimal equalizer parameter based on the NLMS algorithm Calculate the expression 5:
- the value of the parameter ⁇ is also selected in combination with the actual operation of the algorithm to ensure that the convergence speed of the algorithm is maximized when the offset error is small.
- the parameter estimates of the kth equalizer are: ⁇ n , f k , and the parameter estimates of the kth equalizer are categorized. Processed, got i.
- the anti-debt signal processed by the first k equalizers is obtained, and its expression is - /'1 ⁇ 2N-
- step b According to the estimation process of the equalizer parameters in step b, continue to complete the parameter estimation of the k+i, k+2, ..., K equalizers, and after the parameter estimation of the Kth equalizer is completed, the root error value is equalized.
- e (K) is less than the expected root mean square error eo set by the user. The algorithm does not continue to calculate the parameters of the next level equalizer and stops iteration.
- the algorithm will always monitor the magnitude of the equalized root error value e (K) , and compare it with the expected root mean square error set by the user after each iteration is completed. And control the operation of the iterative loop. For example, before the start of the k+i iteration, if e ⁇ k) > Co , then k+1 iterations are continued; if e (k) e 0 , the iteration is stopped.
- the cascaded equalizers in step (3) form a composite equalizer, and the implementation process is as follows:
- the estimated values of the K equalizers obtained by K iteration estimation are -f ( , ..., ,
- the expression of the synthetic equalizer formed by the cascade of these K equalizers is: (1, ' ) . ⁇ (1,2,-- ⁇ ) (1,2, ⁇ - ⁇ ') ⁇ i:i, --K ) ⁇ ,
- step 4 the synthetic equalizer is placed in the signal processing channel to complete the system frequency response equalization operation, which is specific;
- the equalization operation is realized by obtaining a finite impulse response filter of the synthetic equalizer ⁇ ⁇ , assuming that the time domain sequence vector of the input sound source signal is: s - [.?(0) s(l)... ( ⁇ ' ⁇ -1)1 ⁇ '
- the synthesized sound equalizer ⁇ ⁇ processed sound source signal can be expressed as; Two [' (()) ... ⁇ —— 1 )]
- the sound source signal processed by the synthetic equalizer is amplified by the power amplifier and sent to the speaker end, thereby driving the speaker to radiate the sound wave.
- the sound field transfer function from the speaker to the listener position is balanced.
- the peaks and valleys of the transfer function response curve are suppressed, thereby improving the quality of the acoustic playback signal.
- the signal processing channel in the step (4) refers to a channel for performing amplitude adjustment and filtering operation in the digital signal processor to obtain a transmission signal suitable for the output bandwidth requirement of the subsequent stage.
- Another technical solution provided by the present invention is: an online iterative sound reproduction system frequency response equalization device, as shown in FIG.
- the sound source is the sound information to be reproduced by the system:
- a digital signal processor connected to the output end of the sound source, for calculating the parameter estimation values of the equalizers of the multiple iterations, and combining the multi-level equalizer to form a synthetic equalizer, and then adding the synthetic equalizer In the signal processing channel;
- the power amplifier is connected to the output end of the digital signal processor, and ffl is used for power amplification of the equalized signal to drive the speaker to sound;
- the sound source is an analog sound source signal generated by various analog devices, or a digital coded signal generated by various digital devices, or a wireless network transmission signal, and the wireless network transmission signal is transmitted by the wireless transmitting device.
- the incoming broadcast signal is received and demodulated by a wireless receiver to obtain a user-specified sound source signal.
- the analog signal is converted into a system-specified digital input format by means of an analog-to-digital converter; when the sound source is a digital coded signal, the digital coded signal needs to be internal to the digital signal processor. Converted to the system-specified digital input format; When the sound source is a wireless network transmission signal, the signal demodulated by the wireless receiver needs to be converted into a digital input format developed by the system.
- the sound reproduction system has an equalization mode and a normal play mode, and the sound source needs to be set and selected according to two different working modes of the sound reproduction system.
- the sound source is a noise signal generated by a white space sequence or a maximum length sequence (MLS) of a specified bandwidth, and the signal exhibits a flat power spectrum characteristic for use.
- the function of the speaker to the position of the microphone in the free field or the reverberation field is trained; when the sound reproduction system is rotated into the normal play mode, the sound source is the user-specified sound source signal that needs to be played back.
- the sound reproduction system has an equalization mode and a normal play mode, and the digital signal processor needs to perform corresponding processing according to two different working modes in which the sound reproduction system is placed.
- the digital signal processor first processes the noise signal channel signal to complete the amplitude adjustment and filtering operation of the signal, obtain a transmission signal suitable for the output bandwidth requirement of the latter stage, and then send it to the signal.
- Power amplifier side After receiving the feedback signal, the noise source signal and the feedback signal are analyzed and processed, and the parameter estimation values of the equalizers of each level are calculated according to the iterative LMS standard, and then all the equalizers are cascaded to obtain the parameter estimation of the synthetic equalizer.
- the input interfaces of the power amplifier can be divided into two types, namely, a digital input interface and an analog input interface. If the power amplifier has a digital input interface, the digital signal sent from the digital signal processor can be directly amplified by power and sent to the speaker terminal, so the power amplifier is directly connected to the digital signal processor; if the power amplifier only has an analog input
- the interface needs to rely on a digital-to-analog converter to convert the digital signal sent by the digital signal processor into an analog signal, then perform power amplification processing, and finally send it to the speaker terminal, so that there are several connections between the power amplifier and the digital signal processor. Analog converter.
- the speaker is not limited to a single speaker unit, and the speaker may be implemented as a single speaker unit or a speaker array composed of a plurality of speaker units, and the shape of the array may be based on the number of speaker units and actual application requirements. Arrange to form various array shapes suitable for practical application needs.
- the feedback signal receiving module is implemented and operated in accordance with the equalization target of the sound reproduction system. If the sound reproduction system only needs to balance the frequency response characteristics of the circuit system composed of the two parts of the signal processing and the power amplifier driving, the feedback signal receiving module receives and collects the signal of the output of the power amplifier, and sends the collected digital sequence.
- the feedback signal receiving module if the acoustic playback system requires a response fluctuation characteristic after the equalization circuit portion and the speaker portion are coupled, the feedback signal receiving module will collect the microphone received signals at a desired position in the space, and The collected digital sequence is sent to the digital signal processor; if the acoustic playback system requires the response fluctuation characteristic of the equalization circuit portion, the speaker portion and the external environment portion to be coupled together, the feedback signal receiving module will sequentially place the space in the space The microphones at the plurality of location points receive signals for acquisition, and the acquired digital sequences corresponding to the plurality of spatial location points are sent to the digital signal processor.
- the present invention provides an online iterative sound reproduction system frequency response equalization method by increasing the number of cascaded equalizers and increasing each level of equalizers.
- the order of the channel can significantly improve the channel response compensation capability of the synthesized equalizer, so that the overall frequency response curve of the equalized system is more straight, and the synthetic equalizer is approximated to the ideal inverse filter response;
- the online iterative sound reproduction system frequency response equalization method proposed by the present invention can perform parameter estimation and update processing of the equalizer on-line for the environmental change of the sound reproduction system and the performance change of the speaker unit itself.
- the sound reproduction system only needs to be rotated into the equalization mode, and the system automatically emits a noise signal and passes
- the microphone receives and records the feedback signal, and based on the analysis of the noise and feedback signals, the parameter estimation and update of the synthetic equalizer is completed in real time.
- This kind of response compensation method based on online automatic equalization can better meet the needs of practical applications, and also simplifies the equalization operation process, saves the balanced operation time, and its application is more extensive and flexible;
- the multiple iterative equalization method proposed by the present invention can achieve a more balanced equalization process for the entire wideband internal frequency response desired by the user by increasing the number of iterative equalizations, that is, increasing the number of cascaded equalizers. , its ability to balance the frequency response of the low frequency band is significantly better than the traditional equalization method;
- the traditional equalizer parameter estimation method needs to transform the time-frequency domain to obtain the minimum phase response component to realize complex and complicated lock.
- the multiple iterative equalization method proposed by the invention directly analyzes the noise signal and the feedback signal in the time domain, and directly performs the parameter estimation of the equalizer in the time domain, and the signal processing flow thereof
- the hardware implementation is relatively simple;
- the present invention uses an iterative algorithm of LMS or NLMS to estimate the parameter value of the if equalizer.
- This parameter estimation algorithm can be implemented in digital signal processing devices such as DSP and FPGA, and the hardware implementation is simple and the cost is low;
- the present invention generates a synthetic equalizer by cascading a plurality of estimated equalizers. In practical applications, a single synthetic equalizer is used to perform channel equalization operations, which is simple and reliable.
- Figure 1 is a flowchart showing the signal processing of an online iterative sound reproduction system frequency response equalization method and apparatus of the present invention
- Figure 2 is a diagram showing an online iterative sound reproduction system frequency response equalization method and apparatus of the present invention. Schematic diagram of equalization processing of different links;
- FIG. 3 is a schematic diagram of parameter estimation of a multi-stage equalizer of an online iterative sound reproduction system frequency response equalization method and apparatus, wherein s is a sound source, r is a feedback signal, and ⁇ is an equalized sound.
- Source signal FIG. 4 is a schematic diagram showing parameter estimation of the first-order optimal equalizer in the first iteration process of the present invention, wherein r(ii) is an inverse error signal, and white noise ⁇ - ⁇ ) is inverse filtering: Reimbursed sound source signal; flgN
- FIG. 5 is a schematic diagram showing the iteration of the parameter of the i-th optimal equalizer in the first iteration process of the present invention, wherein ⁇ ( ⁇ 1) is the feedback 'is the error signal, and 2 ⁇ is the step size. ⁇ f , is the equalizer parameter;
- FIG. 6 is a schematic diagram of parameter estimation of the k-th optimal equalizer in the k-th iteration process of the present invention, wherein the feedback signal, . ⁇ , is the error signal, is white noise, . Sound source signal compensated for inverse filtering
- Figure ⁇ represents the iterative diagram of the parameter of the k-th optimal equalizer in the kth iteration of the present invention, where / ( for the feedback signal, the error signal, 2 ⁇ 3 ⁇ 4
- Ft (10 i) is the parameter of the device
- Figure 8 is a schematic diagram showing the implementation process of the composite equalizer of the present invention, wherein the input sound source signal, .
- FIG. 9 is a schematic diagram of each component module of an online iterative audio reproduction system frequency response equalization apparatus according to the present invention.
- FIG. 10 is a time-domain waveform diagram of a noise source signal when the system operates in an equalization mode according to an embodiment of the present invention;
- Figure 11 is a diagram showing the waveform of the feedback signal received by the microphone when the system is operated in the equalization mode according to the embodiment of the present invention
- Figure 12 is a diagram showing the frequency response of the system after the system is not applied equalization, equalized by one iteration, and equalized by 10 iterations in the embodiment of the present invention. Curve comparison chart.
- the traditional acoustic reproduction system equalization method is based on the analysis of the system impulse response function, fitting the zero-pole model of these impulse response functions, and then finding the inverse filter response of the system through the pole-zero inversion, thus obtaining the Equalizer parameters for the sound reproduction system.
- the parameter estimation process of these methods relies on the minimum mean square error (LMS) algorithm or the linear estimation coding (LPC) algorithm to calculate the parameters of the equalizer, which is based on the single estimation method.
- LMS minimum mean square error
- LPC linear estimation coding
- the present invention proposes an online iterative sound reproduction system frequency response equalization method and device, by adopting The multiple iterative estimation method calculates the parameter values of multiple cascaded equalizers step by step.
- the synthetic equalizer formed by the cascade of equalizers can better approximate the ideal system inverse filter response, thus reducing the equalization.
- the parameter estimation error of the device ensures that the system frequency response curve has better flatness characteristics after equalization.
- an audio reproduction system frequency response equalization apparatus is constructed, and the main sound source 1, the digital signal processor 2, the power amplifier 3, the speaker 4, and the feedback signal receiving module 5 are composed. .
- the sound source 1 is a white noise signal, and its sampling rate is 23.8 ⁇ , and the number of bits is 16, and its time domain signal waveform is as shown in FIG. 10 .
- the sound source: I is the signal to be reproduced specified by the user.
- the digital signal processor 2 is connected to the output end of the sound source 1, and can be implemented by a DSP or an FPGA as a core processor in hardware implementation.
- the digital signal processor 2 In the equalization mode, the digital signal processor 2 combines the noise signal and the feedback signal to calculate the parameters of the equalizers through multiple iteration estimation algorithms, and cascades all equalizers on the basis of completing the multi-level equalizer estimation.
- a synthetic equalizer is formed and a composite equalizer is implemented using a FiR filter.
- the digital signal processor 2 will perform equalization processing on the reproduced signal using the synthetic equalizer based on the FIR structure.
- the power amplifier 3 is connected to the output of the digital signal processor 2, and performs digital-to-analog conversion and power amplification processing on the digital signal sent from the digital signal processor 2.
- the speaker 4 is connected to the output of the power amplifier 3 to realize electroacoustic conversion and to reproduce an acoustic signal in the air.
- the speaker 4 is a speaker with a diameter of 3, 5 inches, a rated power of 0 watts, and a DC resistance of 4 ohms placed in a closed box.
- the feedback signal receiving module 5 is connected to the output end of the speaker 4. In the equalization mode, the feedback signal receiving module collects a response sequence generated by the noise source and sends it to the digital signal processor 2.
- the sound reproduction system operates in the equalization mode
- the sound source is the white noise signal shown in FIG. 10
- the microphone is placed at 1 meter on the axis of the speaker unit, and the feedback signal recorded by the microphone is
- the waveform is shown in Figure 11. Assume that the order of the equalizers to be estimated is 600, and the number of times of iterative equalization is "0".
- Figure 12 shows a comparison of the system frequency response curves in unbalanced, 1 iterative equalization and 10 iterative equalizations: Comparing these three sets of curves, it can be seen that the system frequency response curve has a very obvious peak in the frequency range of 1,5 ⁇ 4.5 KHz without applying an equalizer; after the first iteration equalization process, the system is The peak value in the frequency range of 1.5KHz ⁇ 4.5KHz has been eliminated, but there is still a small amount of fluctuation in the system frequency response curve in the region near the L5KHz frequency point, and the system frequency response curve in the band of 100Hz ⁇ 200Hz still exists.
- the iterative equalization method proposed by the invention can significantly improve the flatness of the system after the equalization frequency response curve. This shows that the multiple iterative equalization method proposed by the present invention has better equalization effect than the traditional equalization method, and the frequency response curve after equalization will be more straight.
Landscapes
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Otolaryngology (AREA)
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Signal Processing (AREA)
- Circuit For Audible Band Transducer (AREA)
Abstract
一种在线迭代的声重放系统频响均衡方法和装置。该方法包括:(1)将系统旋置均衡模式,设定声源(1)为噪声信号,然后控制系统播放此噪声信号,同时传感器采集和记录反馈信号;(2)结合噪声信号和反馈信号,利用逐级迭代的最小均方误差准则依次估计出各级均衡器的参数;(3)级联各级均衡器,形成合成均衡器;(4)将合成均衡器置于信号处理通道,完成系统频响均衡操作,然后将系统旋置正常播放模式。该装置包括:声源(1)、数字信号处理器(2)、功率放大器(3)、扬声器(4)、反馈信号接收模块(5)。
Description
在线迭代的声重放系统频响均衡方法和装置 技术领域
本发明涉及一种声重放系统的频响均衡方法和装置, 特别涉及一种在线迭代的声重放系 统频响均衡方法和装置。
背景技术
近年来, 随着大规模集成电路和数字信号处理技术的迅速发展, 基于数字信号处理技术 的声重放系统响应均衡问题也逐渐受到国内外多家研究坑构及企业的关注, 并有几家公 司推出了几款带有响应均衡功能的声学产品。 乌普萨拉大学(Uppsa!a University)旗下的 Dirac公司推出了 Dirac HD Sound技术,用干解决自由场环境下扬声器单元的频响曲线峰 谷点均衡, 同时还推出了 Dirac Live技术, 于解决房间内声重放系统的频响曲线峰谷点 均衡。 香港正然传讯有限公司推出了 CONEQ技术, 用于解决房间内扬声器系统的频响 起伏均衡, 该技术利用单个传声器在扬声器单元前方按照蛇形走线路径逐点采集扬声器 系统的脉冲响应数据, 然后通过特定的响应数据处理算法, 实行对扬声器系统的频响均 衡。丹麦林道夫 CLYNGDORF)公司推出了房间均衡技术—— Room Perfect, 该技术利用 阜个传声器在房间内的多个位置点采集扬声器系统的响应数据, 并利 ]¾这些多点响应信 息完成听者位置的响应曲线均衡。美国 KRK公司也推出了房间内声重放系统响应均衡产 品 Ergo (Enhanced Room Geometry Optimization), 该产品也是利用单个传声器在听者 位置四周采集响应数据, 并处理这些响应数据获得均衡器的参数, 完成对听者位置点的 响应均衡。
上述的多款声重放系统均衡产品, 用于均衡自由场内的扬声器单元或者均衡房间内的扬 声器系统, 在自由场内扬声器单元的脉冲响应序列持续时间较短, 脉冲拖尾较小, 其均 衡器的设计较为简单, 但是在房间内扬声器单元的脉冲响应序列持续 间较长, 脉冲拖 尾非常严重, 其均衡器的设计较为复杂, 需要联合多个空间位置点的响应数据进行综合 处理。 这些均衡产品所采 ^均衡算法, 本质上都是通过对采集的时域响应序列迸行分析, 将其分解为最小相位响应部分和全通响应部分, 针对最小相位响应部分直接对其响应的 零极点进行倒置就可以获得其逆滤波器响应的零极点, 从而获得其逆滤波器响应的参数; 对于全通响应部分, 如果其相位响应在期望的频带区域内是恒定值, 不随频率发生变化, 那么这种相位恒定的全通响应部分对重放信号的音色没有明显的影响, 可以忽略不计, 但是如果全通响应部分的相位响应在期望频带内会隨频率发生变化, 那么这种相位响应 随频率变化的全通响应部分会引起重放信号的音色发生明显变化, 需要考虑对其进行均 衡处理。
对自由场及混响场内扬声器系统均衡方法的研究文献较多, 其中一些代表性的研究成 果如下-
文献 1 Stephen T, Neely and Jont B. Allen, "Invertibiiity of a room impulse response ," J.
Acoust. Soc. Am, Vol. 66, No. 1, pp. 165- 169, July 1979.——提出了房间响应中最小相位 响应部分的计算方法, 并对最小相位响应的零极点倒置, 获得了最小相位响应的逆滤波 响应, 基于最小相位部分的逆滤波器参数对房间响应迸行均衡处理, 并通过实验证实了 算法的有效性。
文献 2 Yoichi Haneda , Shoji Makino, and Yutaka Kaneda, " Common acoustical pole axid zero modeling of room transfer functions, " IEEE Transaction on Speech and Audio Processing. Vol 2, No. 2, pp. 320-328, April 1994.——针对房间内多个位置点采集的多组传递函数, 提出了基于共声学极点模型的房间传递函数融合算法, 减少了房间模型的估计参数, 提 高了计算速度。
文献 3 Aki Hanna, Matti aqalainen, Lauri Savioja, Vesa Valimaki, Unto K, Lairie, and jyri Huopaniemi, "Frec!uency- warped signal processing for audio application, " j. Audio Eng. Soc, \¾1. 48, No. 11, pp. 1011-1029, November 2000,——提出基于弯折频率域的响应均衡 方法, 利用线性预测编码 (Linear Predictive Coding—— LPC) 方法 i卜算弯折频率域内的 均衡器参数, 从而提高对低频区域的均衡能力。
传统的声重放系统均衡方法, 都是基于对系统脉冲响应函数进行分析, 拟合这些詠冲响 应函数的零极点模型, 再通过零极点倒置找出系统的逆滤波器响应, 从而获得了声重放 系统的均衡器参数。 这些方法的参数估计过程都是依赖于最小均方误差 (Least Mean Squares—— LMS )算法或者线性预测编码 LPC )算法的阜次估计来计算均衡器的参数, 这种基于单次估 i卜方法所获得的逆滤波器参数与理想的逆滤波器参数之间扔然存在着一 定程度的偏差, 这些偏差将造成均衡后的声重放系统频响曲线在一些频带内仍有较为明 显的峰谷起伏性, 仍未达到较为理想的频响平直特性。 这些均衡后频响曲线的峰谷起伏 特性是由均衡器的参数误差所造成的, 为了削弱均衡后频响曲线的峰谷伏特征, 需要进 一步提高均衡器的参数估计精度, 因此需要寻找更为精确有效的均衡器参数估计方法。 对现有基于单次 LMS或者单次 LPC参数估计算法,在系统均衡器参数估计方面所存在 的一定误差缺陷, 需要考虑采用多次迭代估计的方法, 通过多次迭代操作使所估计的级 联均衡器响应逐歩逼近理想的系统逆滤波器响应, 减少均衡器的参数估计误差, 从而保 证均衡后频响曲线具有更好的平直特性。
发明内容
本发明的目的在于克服基于单次 LMS或者单次 LPC参数估†算法在系统均衡器参数估† 方面所存在的一定误差缺陷, 提供一种在线迭代的声重放系统频响均衡方法和装置。 为了达到上述目的, 本发明采用的技术方案如下: 一种在线迭代的声重放系统频响均衡 方法, 如图 1所示, 包括如下步骤:
( 1 ) 将系统旋置均衡模式, 设定声源为噪声信号, 然后控制系统播放此噪声信号, 同时
传感器采集和记录反馈信号:
(2) 结合噪声信号和反馈信号, 利 ]¾逐级迭代的 LMS准则依次估计出各级均衡器的参 数;
( 3 ) 级联各级均衡器, 形成合成均衡器;
( 4 )将合成均衡器置于信号处理通道, 完成系统频响均衡操作, 然后将系统旋置正常播 放模式。
进一步地, 步骤 (1 ) 中所述噪声信号, 可以为指定带宽的白噪声序列或者最大长度序列 ( Maximum Length Sequence—— MLS ) 所产生的噪声信号, 这种信号呈现平坦的功率谱 特性, 以用于 yii练出自由场或混响场内扬声器到传声器位置点的传递函数。
进一步地, 步骤 (υ 中所述传感器采集和记录反馈信号, 如图 2所示, 这种传感器的类 型是根据声重放系统的均衡目标来确定的; 如果声重放系统仅需要均衡信号处理和功放 驱动两个部分所构成电路系统的频响起伏特性, 这种传感器可以为线缆或模数转换器以 获得经噪声激励所产生的电路处理部分的脉冲响应序列信号; 如果声重放系统需要均衡 电路部分和扬声器部分耦合后的响应起伏特性, 则这种传感器为置于空间某个期望位置 点的传声器, 于记录由嗓声激励所产生的从扬声器到传声器之间对应于声场传播的脉 冲响应序列; 如果声重放系统需要均衡电路部分、 扬声器部分和外部环境部分三个部分 一起稱合后的响应起伏特性, 劑这种传感器为置于空间某一位置点的传声器, 这种传声 器需要改变所处空间位置, 依次记录多个空间位置点上 ώ噪声激励所产生的从扬声器到 传声器之间对应于声场传播的脉冲响应序列。
进一步地, 歩骤 (2) 中所述利 ¾逐级迭代的 LMS准则依次估 i卜出各级均衡器的参数, 这种基于逐级迭代的 LMS准则的各级均衡器参数估计过程如下:
a. 假设输入噪声源信号的时域序列矢量为:
其中, Ν 是噪声源信号时域离散序列的采样点数; 假设声重放系统的时域脉冲响应序列 表达式:
其中, M为系统时域脉冲响应的序列长度;
= sw * h
其中 " 代表两个序列矢量之间迸行卷积操作; 如图 3所示, 在第 1次迭代时, 假定待 求的第 1个均衡器的时域脉冲响应序列的长度为 L(L〉M) , 那么该响应序列矢量可以表示 为:
反馈信号 Γ经由待求的第 1个均衡器/ ^处理后, 得到经第 1次均衡后的反馈信号表达式
结合声源信号 sw和第 1 次均衡后的反馈信号 u, 根
据最小均方误差准则,计算出使声源信号 sw和均衡后反馈信号 之间均方误差取最 Φ值
'
= arg min i?{||r ~^J| } 如图 4所示, 根据 LMS算法可以计算出最优均衡器参数矢量的估计值, 其最优均衡器参 数的计算过程是通过基于 LMS的迭代算法来实现的, 在第 n次迭代过程中, 第 1个最优 均衡器参数的计算表达式如下- r^'(n) ^ 2^ J (n) ' r("― ) (")二 (" A) )(") i g.i 0? + 1) - heqjin) + 2μ ' e^(n) . ri ? - /), 0 < < L - 1 '
其中 Δ代表声源信号 Sw (n) 的延迟量; 根据上述表达式可知, 第 : I个最优均衡器的参数 迭代过程, 其中 μ为歩长, 其取值满足如下关系:
1
0 < μ <
t K)
其中 tr (R) 为矩阵 R的迹, R E(r.r , 是矢量「的共轭转置矢量, ίίί!「是由传感 器接收的反馈信号所成的长度为 L的时域序列矢量, 在第 i个釆样时刻反馈信号 r ω的 矢量表达式为: r(i) - \r(i) r(i— 1)… r(i -L + 1)]
歩长 μ按照下面关系来选取:
1
0 < μ <
λ,.
其中 A 为矩阵 R的最大特征值;
max
步长因子 μ的实际取值大小要结合 LMS算法的实际运行情况来选择, 以保证在失调误差 较小的前提下算法具有较好的收敛速度。
为了提高 LMS算法的收敛速度,最优均衡器参数的计算也可以采用归一化的 LMS算法, 即 NLMS(Normaiized Least Mean Squares)算法, 则在第 n次迭代过程中 , 基干 NLM:S算 法的第 1个最优均衡器参数的计算表达式如下: req '(η) 2^ (η) ' r(n .
keq (ft + 1) heqj ί ?) + ―
γ -f- r(i
其中 μ为步长因子, Υ为小正常数,
r(n) Γ(Π ~1) 'Γ(Π ~Γ}
ϋ
对于 NLMS算法, 参数 μ和 Y的取值也要结合算法的实际运行情况来选择, 以保证在失 调误差较小的情况下, 尽量提高算法的收敛速度。
其中, )「是矢量 £^的转置矢量; 反馈信号 Γ经 ώ第 ί个均衡器 处理后, 得到经第 1 次均衡后的反馈信号表达式为: '(。 )… 1)] 结合第 1次均衡后的反馈信号 )和声源信号 sw, 我们可以^算出经第〗级均衡器处理后 eq
的反馈信号与声源信号之间的均方根误差值 其表 i
max (1)、 max(,s'¾[)
N
b. 按照步骤 a中均衡器参数的估 i卜过程, 继续完成第 1、 2、 …、 k L个均衡器的参数估 计, 假设在完成第 k -】个均衡器参数估计的基础上, 如图 6所示, 第 k个均衡器参数的估 计过程如下:
经前 k- 1个均衡器滤波处理后的反馈信号 n再经由待求的第 k个均衡器 U 处理后,得 到经前 k个均衡器处理后的反馈信号表达式为:
结合声源信号 sw和经 k个均衡器处理后的反馈信号 ,根据最小均方误差准剣,计算出 使声源信号 sw和均衡后反馈信号 之间均方误差取最小值时的均衡器参数矢量,作为第 k次迭代 ^设计的第 k个最优均衡器参数矢量的估计值, 其表达式为-
如图 6所示, 根据: LMS算法可以计算出最优均衡器参数矢量的估计值, 其最优均衡器参 数的计算过程是通过基于 LMS的迭代算法来实现的, 在第 I次迭代过程中, 第 k个最优 均衡器参数的计算表达式如下:
、 ·) w
heqj (η -ί- 1) - h qJ (") - 2/i - (") - r^ ) (n - ), ()</</. - 1 '
根据上述表达式可知, 第 k个均衡器参数的迭代过程如图 7所示, 其中步长 μ的取值满
足如下关系:
是矢量 的共轭转置矢量, 而 是经前
0 <·. μ < k— 1个均衡器滤波处理后的反馈信号所组成的长度为 L的时域序列矢量, ¾ 采样时 刻经前 k- i个最优均衡器滤波处理后的反馈信号, 其时域的矢量表达式为:
另外, 步长 μ按照下面关系来选取:
1 '
0 < μ < 其中 A raax为矩阵 R的最大特征值; 步长因子 μ的实际取值大小要结合 LMS算法的实际 运行情况来选择, 以保证在失调误差较小的前提下算法具有较好的收敛速度。
/½,,(« + 1) - : n) + - μ
对于 NLMS算法, 参数 μ和 的取值也要结合算法的实际运行情况来选择, 以保证在失 调误差较小的情况下, 尽量提高算法的收敛速度。
假定 ώ LMS算法或者 NLMS算法经 N«次迭代计算后算法趋于收敛, 此时第 k个均衡器 的参数估计值为: ^n 、 fk、 对第 k个均衡器的参数估计值 进行归 化处理, 得至 i.
^处理后, 得到经前 k个均衡器处理后的反债信号, 其表达式为- /'½N-
(k)
M
c. 按照步骤 b中均衡器参数的估计过程, 继续完成第 k+i、 k+2, …、 K个均衡器的参数 估计,在完成第 K个均衡器的参数估计后,均衡根误差值 e(K)小于 户设定的期望均方根 误差 eo, 算法不再继续†算下一级均衡器的参数, 停止迭代。
更进一步地, 在估^均衡器参数的迭代计算过程中, 算法会一直监视均衡根误差值 e(K) 的大小, 在每次迭代完成后都会与用户设定的期望均方根误差 进行比较, 并控制迭代 循环的运行。 如, 在第 k+i次迭代开始之前, 如果 e<k)> Co , 那么继续执行 k+1次迭代; 如果 e(k) e0, 则停止迭代。
进一步地, 歩骤 (3 ) 中的级联各级均衡器, 形成合成均衡器, 其实现过程如下: 由 K次迭代估计所获得的这 K个均衡器的参数估计值分别为- f ( 、 …、 , 由这 K个均衡器级联所形成的合成均衡器 ·· 的表达式为: (1, ' ) .^(1,2,--· ) (1,2,·-Α') ^{i:i, --K ) ν,
其中 代表时域矢量序列之间的卷积操作。
进一步地, 歩骤 4) 中的将合成均衡器置于信号处理通道, 完成系统频响均衡操作, 其 具体;
在获得合成均衡器 ^Α ·^的基础 利 ]¾有限脉冲响应滤波器来实现均衡操作, 假设输 入声源信号的时域序列矢量为: s - [.?(0) s(l)… Λ(Λ' ~-1)1γ '
经合成均衡器 ^ ··· 处理后的声源信号可以表示为;
二 [' (()) … ^^——1)]
经合成均衡器处理后的声源信号, 通过功率放大器放大后送至扬声器端, 从而驱动扬声 器辐射声波, 通过这种均衡处理操作, 从扬声器到听者位置点处的声场传递函数得到了 均衡, 其传递函数响应曲线的峰谷得到了抑制, 从而提高了声重放信号的质量。
进一步地, 歩骤 (4) 中的信号处理通道, 是指在数字信号处理器内完成信号的幅度调整 以及滤波操作, 以获得适合后级输出带宽要求的传输信号的通道。
本发明所提供的另一技术方案为: 一种在线迭代的声重放系统频响均衡装置, 如图 ί 所 示' 它包括:
声源, 是系统待重放的声信息:
数字信号处理器, 与声源的输出端相连接, 用于^算多次迭代的各级均衡器的参数估^ 值, 并联合多级均衡器形成合成均衡器, 然后将合成均衡器加入到信号处理通道中; 功率放大器, 与数字信号处理器的输出端相连接, ffl于对均衡处理后的信号迸行功率放 大, 以驱动扬声器发声;
扬声器, 与功率放大器的输出端相连接, 用于电声转换以便将声源信号重放到空气中: 反馈信号接收模块, 与扬声器的输出端相连接, 用于采集和记录重放到空气中的声信号。 进一歩地, 声源为来自于各种模拟装置所产生的模拟声源信号, 或者为各种数字装置所 产生的数字编码信号, 或者为无线网络传输信号, 无线网络传输信号为无线发射装置传 送来的广播信号并通过无线接收器迸行接收和解调获得用户指定的声源信号。 当声源为 模拟声源信号时, 需要借助模数转换器, 将模拟信号转换为系统指定的数字输入格式; 当声源为数字编码信号时, 需要在数字信号处理器内部将该数字编码信号转换为系统指 定的数字输入格式; 当声源为无线网络传输信号时, 需要将无线接收器解调的信号转换 为系统制定的数字输入格式。
进一歩地, 声重放系统具有均衡工作模式和正常播放模式, 声源需要根据声重放系统的 两种不同的工作模式迸行设置选择。 当声重放系统置于均衡工作模式^, 声源为指定带 宽的白嗓声序列或者最大长度序列 (Maximum Length Sequence MLS ) 所产生的噪声 信号, 这种信号呈现平坦的功率谱特性, 以用于 ^练出自由场或混响场内扬声器到传声 器位置点的传递函数; 当声重放系统旋置于正常播放工作模式时, 声源为用户指定的需 要重放的声源信号。
更进一步地, 声重放系统具有均衡工作模式和正常播放模式, 数字信号处理器需要根据 声重放系统所置于的两种不同的工作模式进行相应的处理工作。 当声重放系统置于均衡 工作模式时, 数字信号处理器首先将噪声信号迸行通道信号处理, 以完成信号的幅度调 整以及滤波操作, 获得适合后级输出带宽要求的发射信号, 然后再送至功率放大器端,
在接收到反馈信号后会对噪声源信号和反馈信号进行分析处理, 按照迭代 LMS准劑, † 算出各级均衡器的参数估计值, 然后级联所有的均衡器, 获得合成均衡器的参数估计值, 最后用合成均衡器的参数估计值更新 FIR滤波器的相应的系数值, 丛而将合成均衡器加 入到信号处理通道中; 当声重放系统置于正常播放工作模式时, 数字信号处理器将按照 合成均衡器的参数估计值对用户指定的输入信号进行均衡处理, 并将均衡后信号送至功 放输入端。
进一歩地, 功率放大器的输入接口可以分为两种类型, 即数字输入接口和模拟输入接口。 如果功率放大器具有数字输入接口, 則可以直接对数字信号处理器送来的数字信号进行 功率放大处理后再送至扬声器端, 因此功率放大器直接与数字信号处理器相连接; 如果 功率放大器仅具有模拟输入接口, 则需要依靠数模转换器, 将数字信号处理器送来的数 字信号转换为模拟信号后再进行功率放大处理, 最后送至扬声器端, 因此功率放大器与 数字信号处理器之间连接有数模转换器。
进一步地, 扬声器并不局限于单一的扬声器单元, 扬声器的实现形式可以为单个扬声器 单元, 也可以为多个扬声器单元组成的扬声器阵列, 并且该阵列的形状可以根据扬声器 阜元数量和实际应用需求进行排列, 组成适合于实际应用需求的各种阵列形状。
进一歩地, 反馈信号接收模块, 其实现和工作方式需要根据声重放系统的均衡目标来确 定。 如果声重放系统仅需要均衡信号处理和功放驱动两个部分所构成电路系统的频响起 伏特性, 反馈信号接收模块则对功放输出端的信号迸行接收和采集, 并将采集的数字序 列送入数字信号处理器中: 如果声重放系统需要均衡电路部分和扬声器部分耦合后的响 应起伏特性, 则反馈信号接收模块将对置于空间某个期望位置点的传声器接收信号进行 釆集, 并将采集的数字序列送入数字信号处理器中; 如果声重放系统需要均衡电路部分、 扬声器部分和外部环境部分三个部分一起耦合后的响应起伏特性, 则反馈信号接收模块 将对依次置于空间多个位置点处的传声器接收信号迸行采集, 并将采集的对应于多个空 间位置点的数字序列送至数字信号处理器中。
与现有技术相比, 本发明的优点在于-
1. 与传统的基于单次 LMS或者单次 LPC参数估计算法相比, 本发明所提出的在线迭代 的声重放系统频响均衡方法, 通过增加级联均衡器的数量和增加每级均衡器的阶数, 可 以明显提升合成均衡器的通道响应补偿能力, 使均衡后的系统整体频响曲线更为平直, 并使合成均衡器逼近于理想的逆滤波器响应;
2. 本发明所提出的在线迭代的声重放系统频响均衡方法, 对声重放系统所处的环境变化 和扬声器单元本身的性能变化, 能够在线完成均衡器的参数估计和更新处理。 用户在更 改声重放系统的扬声器单元及安装箱体特性或者更改声重放系统所放置的空间环境时, 仅需要将声重放系统旋置均衡模式, 系统会自动的发出噪声信号, 并通过传声器接收和 记录反馈信号, 基于噪声和反馈信号的分析, 实时完成合成均衡器的参数估计和更新任
务。 这种基于在线自动均衡的响应补偿方式, 更能满足实际应用需求, 同时也简化了均 衡操作流程, 节省了均衡操作时间, 其应用更为广泛和灵活;
3. 本发明所提出的多次迭代均衡方法, 能够通过增加迭代均衡的次数, 即增加级联均衡 器的数量, 来实现对用户期望的整个宽频带内频响进行更为平直的均衡处理, 其对于低 频带内频响的均衡能力要明显优于传统的均衡方法;
4. 传统的均衡器参数估计方法, 需要通过时频域的变换以获取最小相位响应分量, 实现 复杂繁锁。 与传统的均衡器参数估计方法相比, 本发明所提出的多次迭代均衡方法, 直 接在时域内分析噪声信号和反馈信号, 并直接在时域内完成均衡器的参数估计, 其信号 处理流程及硬件实现较为简单;
5. 本发明采用 LMS或者 NLMS的迭代算法来估 if均衡器的参数值, 这种参数估 i卜算法 完全可以在 DSP和 FPGA等数字信号处理器件中实现, 硬件实现简单、 成本较低;
6. 本发明通过对多个已估†的均衡器进行级联生成単一的合成均衡器, 在实际应用中, 使用单个合成均衡器来执行通道的均衡操作, 实现简单可靠。
跗图说明
图 i表示本发明的一种在线迭代的声重放系统频响均衡方法和装置的信号处理流程图; 图 2表示本发明的一种在线迭代的声重放系统频响均衡方法和装置的三种不同环节的均 衡处理示意图;
图 3 表示本发明的一种在线迭代的声重放系统频响均衡方法和装置的多级均衡器的参数 估计示意图, 其中 s为声源, r为反馈信号, ^为经均衡处理后的声源信号; 图 4表示本发明在第 1次迭代过程中第 1级最优均衡器的参数估计示意图,其中 r(ii)为反 为误差信号, 为白噪声 ^ -κ)为逆滤波:: 偿后的声源信号; flgN
图 5表示本发明在第 1次迭代过程中第 i级最优均衡器的参数迭代示意图, 其中 Γ(η- 1) 为反馈' 为误差信号, 2 μ为步长, ? ω f , 为均衡器参数; 图 6表示本发明在第 k次迭代过程中第 k级最优均衡器的参数估计示意图, 其中 为反馈信号, .^ 、为误差信号, 为白噪声, . ι 为逆滤波补偿后的声源信
图 Ί表示本发明在第 k次迭代过程中第 k级最优均衡器的参数迭代示意图, 其中 /( 为反馈信号, 为误差信号, 2 μ ¾
ft ("十 i)为聰器参数;
图 8表示本发明的合成均衡器的实现过程示意图,其中 为输入声源信号, . ^η为
s * hgN
经合成均衡器处理后的输入信号;
图 9表示本发明的一种在线迭代的声重放系统频响均衡装置的各组成模块的示意图; 图 10表示本发明实施例中系统工作于均衡模式时噪声源信号的时域波形图;
图 11表示本发明实施倒中系统工作于均衡模式时传声器接收的反馈信号波形图; 图 12 表示本发明实施例中系统未施加均衡、 经 1次迭代均衡和经 10次迭代均衡后系统 频响曲线对比图。
其中图中标号为:
声源; 2、 数字信号处理器; 3、 功率放大器; 4、 扬声器; 5、 反馈信号接收模块。 具体实施方式
下面结合附图对本发明的较佳实施 ^进行详细阐述, 以使本发明的优点和特征能更易于 被本领域技术人员理解, 从而对本发明的保护范围做出更为清楚明确的界定。
目前, 传统的声重放系统均衡方法, 都是基于对系统脉冲响应函数进行分析, 拟合这些 脉冲响应函数的零极点模型, 再通过零极点倒置找出系统的逆滤波器响应, 从而获得了 声重放系统的均衡器参数。 这些方法的参数估†过程都是依赖于最小均方误差 (LMS ) 算法或者线性预测编码(LPC)算法的单次估†来计算均衡器的参数, 这种基于单次估计 方法所获得的逆滤波器参数与理想的逆滤波器参数之间仍然存在着一定程度的偏差, 这 些偏差将造成均衡后的声重放系统频响曲线在一些频带内仍有较为明显的峰谷起伏性, 扔未达到较为理想的频响平直特性。为了克服基于单次 LMS或者单次 LPC参数估计算法, 在系统均衡器参数估计方面所存在的一定误差缺陷, 本发明提出了一种在线迭代的声重 放系统频响均衡方法和装置, 通过采用多次迭代估计方法逐级计算出多个级联均衡器的 参数值, 这些各级均衡器级联所形成的合成均衡器能够更好的逼近于理想的系统逆滤波 器响应, 从而减少了均衡器的参数估计误差, 保证了均衡后系统频响曲线具有更好的平 直特性。本发明通过增加级联均衡器数量,可以明显提升均衡器的响应均衡能力,同时通过 在线自动均衡,简化了均衡操作流程, 其应用场景更为广泛和灵活。
如图 9所示, 制作一个依据本发明的在线迭代的声重放系统频响均衡装置, 其主体 声 源 1、 数字信号处理器 2、 功率放大器 3、 扬声器 4、 反馈信号接收模块 5等组成。
声源 1 , 当系统旋置均衡模式时声源 1为白噪声信号, 其采样率为 23.8ΚΉΖ, 比特位数为 16, 其时域信号波形如图 10所示。 当系统旋置正常播放模式时, 声源 : I为 ]¾户指定的待 重放信号。
数字信号处理器 2,与所述声源 1的输出端相连接,在硬件实现上可以由 DSP或者 FPGA 作为核心处理器来实现。 在均衡模式下, 数字信号处理器 2将联合噪声信号和反馈信号, 通过多次迭代估^算法^算出各级均衡器的参数, 在完成多级均衡器估计的基础上, 级 联所有均衡器形成合成均衡器, 并利用 FiR滤波器来实现合成均衡器。 在正常播放模式 下, 数字信号处理器 2将利用基于 FIR结构的合成均衡器对待重放信号进行均衡处理。
功率放大器 3, 与所述数字信号处理器 2的输出端相连接, 对数字信号处理器 2送入的数 字信号进行数模变换和功率放大处理。
扬声器 4, 与所述功率放大器 3的输出端相连接, 实现电声变换, 在空气中重放声信号。 扬声器 4为口径为 3,5寸、额定功率为 0 瓦、直流电阻为 4 欧姆、 置于封闭箱体内的扬 声器。
反馈信号接收模块 5, 与所述扬声器 4的输出端相连接, 在均衡模式下, 反馈信号接收模 块将采集 噪声源所激励产生的响应序列, 并送至数字信号处理器 2中。
实施例:
在本实施例中, 假设声重放系统工作于均衡模式下, 扬声器为声源为图 10所示的白噪声 信号, 传声器置于扬声器单元轴线上 1 米处, 传声器所记录的反馈信号^域波形如图 11所示。 假设待估计的各级均衡器的阶数均为 600, 设置迭代均衡的次数为】 0。
图 12给出了在未均衡、 经 1次迭代均衡和经 10次迭代均衡:三种情况下, 系统频响曲线 的对比图。对比这三组曲线可以看出,在未施加均衡器的情况下系统频响曲线在 1,5ΚΗζ〜 4.5 KHz 的频带范围内存在着非常明显的峰值; 在第 1 次迭代均衡处理后, 系统在在 1.5KHz〜4.5KHz的频带范围内的峰值已经得到了消除, 但是在 L5KHz频点 近的区域 内系统频响曲线仍有小量的起伏, 同时在 100Hz〜200Hz的频带 系统频响曲线仍有较 大程度的起伏; 在经 10次迭代均衡处理后, 系统在 1.5ΚΉζ频点附近区域内的少量起伏 已经得到了消除, 同时在 100ΗΖ〜200Ηζ频带内的幅度峰值也得到了较大程度的抑制。对 比第 : 次迭代均衡和第 10次迭代均衡处理后的频响曲线,可以看出:通过增加迭代次数, 本发明所提出的迭代均衡方法, 能够明显改善系统均衡后频响曲线的平直程度, 这说明 本发明所提出的多次迭代均衡方法, 较传统均衡方法相比, 具有更好的均衡效果, 经其 均衡后的频响曲线将会更为平直。
上述实施^只为说明本发明的技术构思及特点, 其目的在于让熟悉此项技术的人士能够 了解本发明的内容并据以实施, 并不能以此限制本发明的保护范围。 凡根据本发明精神 实质所作的等效变化或修饰, 都应涵盖在本发明的保护范围之内。
Claims
1. 一种在线迭代的声重放系统频响均衡方法, 包括如下步骤:
( 1 ) 将系统旋置均衡模式, 设定声源为噪声信号, 控制系统播放此噪声信号, 同时通过 传感器采集和记录反馈信号:
( 2 ) 结合噪声信号和反馈信号, 利用逐级迭代的最小均方误差准则依次计算出各级均衡 器的参数;
( 3 ) 级联各级均衡器, 形成合成均衡器;
( 4 )将合成均衡器置于信号处理通道, 完成系统频响均衡操作, 然后将系统旋置正常播 放模式。
2. 根据权利要求 1所述的在线迭代的声重放系统频响均衡方法, 其特征在于: 步骤 ( ) 中的噪声信号, 为指定带宽的白噪声序列或者最大长度序列所产生的噪声信号。
3. 根据权利要求 i所述的在线迭代的声重放系统频响均衡方法, 其特征在于: 步骤 (2 ) 中的利用逐级迭代的最小均方误差准则依次估计出各级均衡器的参数, 这种基于逐级迭 代的最小均方误差准则的各级均衡器参数估计过程如下- a. 假设输入噪声源信号的时域序列矢量为-
其中, Ν 是嗓声源信号时域离散序列的采样点数; 假设声重放系统的时域脉冲响应序列 表达式为:
h = \ K "' U
其中, M为系统时域脉冲响应的序列长度; 假设传感器采集的反馈信号的时域序列为: r - [r(0) rfl) · · · r( V 1)]】, 其中 代表两个序列矢量之间进行卷积操作; 在第 1次迭代 , 假定待求的第 1个均 衡器的时域脉冲响应序列的长度为 L(L>M), 该响应序列矢量则表示为:
反馈信号「'经由待求的第 1个均衡器 (1¾处理后, 得到经第 i次均衡后的反馈信号表达式 为:
(0) … ! D' (N— i)]r '
™ r * h'、 }
结合声源信号 sw和第 1次均衡后的反馈信号 (n , 根据最小均方误差准则, 计算出使声源
信号 sw和均衡后反馈信号 ω之间均方误差取最小值^的均衡器参数矢量,作为第 1次迭 代时设计的第 1个最优均衡器参数矢量的估计值, 其表达式为-
根据最小均方误差算法计算出最优均衡器参数矢量的估†值, 其最优均衡器参数的计算 过程是通过基于最小均方误差的迭代算法来实现的, 在第 11次迭代过程中, 第 1个最优 均衡器参数的计算表达式如下:
heqj (n + 1)- heqJ (") + 2μ - e^! (n) · r(n -/), 0</<I~-l
其中 Δ代表声源信号 Sw (n) 的延迟量; 根据上述表达式—可知, 第 1个最优均衡器的参数 迭代过程, 其中 μ为步长, 其取值满足如下关系-
0 < μ <
tr(K)
其中 tr (R) 为矩阵 R的迹, R E r,rHy 是矢量 Γ的共轭转置矢量, 而 Γ是由传感 器接收的反馈信号所成的长度为 L的时域序列矢量, 在第 i个采样 ^刻反馈信号 r (i)的 矢量表达式为; r(i) - [r( r(i - 1) · .. r(i - L + 1)]1
歩长 μ按照下面关系来选取; 为矩阵 R的最大特征值; 假定由最小均方误差算法经 次迭代计算后算法趋于收敛, 此时第 1个均衡器的参数 1ό 计值为: fW ...f£(" ? « f., 对第 1个均衡器的参数估计值 进行归一化处理,
其中, '「是矢量 的转置矢量; 反馈信号 Γ经由第 1个均衡器^ υ处理后, 得到经第 1次均衡后的反馈信号表达式为; ; =[ (ο) Κ ν ι)Γ
结合第〗次均衡后的反馈信号 和声源信号 sw, 计算出经第〗级均衡器处理后的反馈信 eq
b. 按照歩骤 a中均衡器参数的估计过程, 继续完成第 1、 2、 …、 i个均衡器的参数估 计, 假设在完成第 k-1个均衡器参数估计的基础上, 第 k个均衡器参数的估计过程如下: 经前 k- i个均衡器滤波处理后的反馈信号 表示为-
(k -\ )
- … - ι)]Γ
经前 k- ί个均衡器滤波处理后的反馈信号 再经由待求的第 k个均衡器^处理后,得 到经前 k个均衡器处理后的反馈信号表达式为:
结合声源信号 sw和经 k个均衡器处理后的反馈信号 rW ,根据最小均方误差准则,计算出 使声源信号 SW和均衡后反馈信号 W之间均方误差取最小值时的均衡器参数矢量,作为第 k次迭代时设计的第 k个最优均衡器参数矢量的估计值, 其表达式为:
、 )
ί i)
heg hegj («) + 2μ - (jl) - ("― /), 0 < / < L ~ 1 '
根据上述表达式可知, 第 k个均衡器参数的迭代过程, 其中步长 μ的取值满足如下关系:
1
0< μ<
tr(R)
- .^ -r>H, , 是矢量 的共轭转置矢量, 而 (m是经前 k -〗 个均衡器滤
Λ - "eg ) 'eq 'eg f eq
波处理后的反馈信号所组成的长度为 L的时域序列矢量,在第 ^个采样时刻经前 k- 1个最 优均衡器滤波处理后的反馈信号, 其时域的矢量表达式为-
C( - °( … -〖)(卜 i)]r ;
歩长 μ按照下面关系来选取; 1
0 < μ < 其中 λ max为矩阵 R的最大特征值:
得到; gN - [ gNJ) hgN "· hgN,L-i ]
hi 其中, 是矢量 £W的转置矢量; 经前 k- 1个均衡器滤波处理后的反馈信号^^再经由
第 k个均衡器 处理后, 得到经前 k个均衡器处理后的反馈信号, 其表达式为:
n -i (O) 1)]
4. 根据权利要求 3所述的在线迭代的声重放系统频响均衡方法, 其特征在于: 在估计均 衡器参数的迭代计算过程中, 算法一直监视均衡根误差值 e 的大小, 在每次迭代完成后 都会与用户设定的期望均方根误差 efi进行比较, 并控制迭代循环的运行, 在下一次迭代 开始之前, 如果 e(k)〉ee, 那么继续执行下一次迭代, 如果 剣停止迭代。
5. 根据权利要求 3所述的在线迭代的声重放系统频响均衡方法, 其特征在于: 在步骤 a 中, 最优均衡器参数的计算采用归一化的最小均方误差算法, 在第 11次迭代过程中, 基 于归一化的最小均方误差算法的第 1个最优均衡器参数的计算表达式如下-
- )
(n)-r(n -
— /), 0< < "1
其中 μ为歩长因子, Υ为小正常数,
: ^ r(«~ )-r(«" )
6. 根据权利要求 1所述的在线迭代的声重放系统频响均衡方法, 其特征在于: 在所述的 步骤 a 中, 所述的最优均衡器参数的计算还可以采用归一化的最小均方误差算法, 則在 第 ίΐ次迭代过程中, 基于归一化的最小均方误差算法的第 k个最优均衡器参数的计算表 达式如下-
h q,l {n H- I) · heq (n) + ·
γ -\- ) 其中 μ为步长因子, Υ为小正常数,
7. 根据权利要求 ί所述的在线迭代的声重放系统频响均衡方法, 其特证在于: 歩骤 (3 ) 中的级联各级均衡器, 形成合成均衡器, 萁实现过程如下:
8. 根据权利要求 1所述的在线迭代的声重放系统频响均衡方法, 其特征在于: 步骤 (4) 中的将合成均衡器置于信号处理通道, 完成系统频响均衡操作, 其具体实现如下: 在获得合成均衡器 的基础上, 利用有限脉冲响应滤波器来实现均衡操作, 假设输 入声源信号的时域序列矢量为:
经合成均衡器 i处理后的声源信号可以表示为:
经合成均衡器处理后的声源信号, 通过功率放大器放大后送至扬声器端, 驱动扬声器辐 射声波。
9. 根据权利要求 1所述的在线迭代的声重放系统频响均衡装置, 其特征在于: 歩骤 (4) 中的信号处理通道, 是指在数字信号处理器内完成信号的幅度调整以及滤波操作, 获得 适合后级输出带宽要求的传输信号的通道。
10. 一种在线迭代的声重放系统频响均衡装置, 其特征在于: 它包括声源(1 )、 与所述声 源 (1 ) 的输出端相连接并用于计算多次迭代的各级均衡器的参数估计值并联合多级均衡 器形成合成均衡器然后将合成均衡器加入到信号处理通道中的数字信号处理器 (2)、 与 所述数字信号处理器(2 ) 的输出端相连接并用于对均衡处理后的信号进行功率放大以驱 动扬声器发声的功率放大器 (3 )、 与所述功率放大器 (3 ) 的输出端相连接并用于电声转 换以便将声源信号重放到空气中的扬声器 (4)、 与所述扬声器 (4) 的输出端相连接并用 于采集和记录重放到空气中的声信号的反馈信号接收模块 (5 ), 所述的声源是系统待重 放的声信息。
11. 根据权利要求 10所述的在线迭代的声重放系统频响均衡装置,其特征在于:声源(1 ) 为来自于各种模拟装置所产生的模拟声源信号, 或者为各种数字装置所产生的数字编码 信号, 或者为无线网络传输信号, 无线网络传输信号为无线发射装置传送来的广播信号 并通过无线接收器进行接收和解调获得用户指定的声源信号。
12. 根据权利要求 10所述的在线迭代的声重放系统频响均衡装置, 其特征在于: 声重放 系统具有均衡工作模式和正常播放工作模式, 当声重放系统在均衡工作模式时, 声源(1 ) 为指定带宽的白噪声序列或者最大长度序列所产生的噪声信号, 当声重放系统旋在正常 播放工作模式时, 声源 (1 ) 为 ffl户指定的需要重放的声源信号。
13. 根据权利要求 10所述的在线迭代的声重放系统频响均衡装置, 其特征在于; 功率放 大器 (3 ) 具有数字输入接口, 功率放大器 (3 ) 直接与数字信号处理器 (2) 连接。
14. 根据权利要求 10所述的在线迭代的声重放系统频响均衡装置, 其特征在于: 功率放 大器 (3 ) 具有模拟输入接口, 功率放大器 (3 ) 与数字信号处理器 (2) 之间连接有将数 字信号处理器 (2) 送来的数字信号转换为模拟信号的数模转换器。
】5, 根据权利要求 10所述的在线迭代的声重放系统频响均衡装置, 其特征在于; 扬声器 (4) 为单个扬声器单元、 或者为多个扬声器单元组成的扬声器阵列。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210388256.5A CN102883243B (zh) | 2012-10-15 | 2012-10-15 | 在线迭代的声重放系统频响均衡方法 |
CN201210388256.5 | 2012-10-15 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2014059890A1 true WO2014059890A1 (zh) | 2014-04-24 |
Family
ID=47484349
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2013/084956 WO2014059890A1 (zh) | 2012-10-15 | 2013-10-10 | 在线迭代的声重放系统频响均衡方法和装置 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN102883243B (zh) |
WO (1) | WO2014059890A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115455745A (zh) * | 2022-11-09 | 2022-12-09 | 深圳飞骧科技股份有限公司 | 自适应频点采样的扫频方法、系统及相关设备 |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102883243B (zh) * | 2012-10-15 | 2015-03-25 | 苏州上声电子有限公司 | 在线迭代的声重放系统频响均衡方法 |
CN103220597B (zh) * | 2013-03-29 | 2014-07-23 | 苏州上声电子有限公司 | 车内声场均衡装置 |
CN105610748B (zh) * | 2014-11-20 | 2018-11-16 | 中国航空工业集团公司雷华电子技术研究所 | 一种频率分段的通道均衡方法 |
CN107071680B (zh) * | 2017-04-19 | 2019-09-13 | 歌尔科技有限公司 | 一种声学产品的调音方法和装置 |
CN111836165A (zh) * | 2020-07-10 | 2020-10-27 | 深圳市昂思科技有限公司 | 一种有源降噪系统中电声器件频率响应曲线的补偿方法 |
CN114390402B (zh) * | 2022-01-04 | 2024-04-26 | 杭州老板电器股份有限公司 | 一种吸油烟机的音频注入控制方法、装置及吸油烟机 |
CN117896219B (zh) * | 2024-03-18 | 2024-09-17 | 中国民航大学 | 一种基于ssa寻优的lms均衡优化方法、设备及介质 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101155438A (zh) * | 2006-09-26 | 2008-04-02 | 张秀丽 | 音频设备的频率响应自适应均衡方法 |
US8098848B2 (en) * | 2006-05-19 | 2012-01-17 | Nuance Communications, Inc. | System for equalizing an acoustic signal |
CN102447446A (zh) * | 2011-12-09 | 2012-05-09 | 苏州上声电子有限公司 | 基于振动元件运动状态反馈的扬声器频响均衡方法和装置 |
CN102883243A (zh) * | 2012-10-15 | 2013-01-16 | 苏州上声电子有限公司 | 在线迭代的声重放系统频响均衡方法和装置 |
CN102903367A (zh) * | 2012-10-15 | 2013-01-30 | 苏州上声电子有限公司 | 离线迭代的声重放系统频响均衡方法和装置 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6111958A (en) * | 1997-03-21 | 2000-08-29 | Euphonics, Incorporated | Audio spatial enhancement apparatus and methods |
US8121312B2 (en) * | 2006-03-14 | 2012-02-21 | Harman International Industries, Incorporated | Wide-band equalization system |
-
2012
- 2012-10-15 CN CN201210388256.5A patent/CN102883243B/zh active Active
-
2013
- 2013-10-10 WO PCT/CN2013/084956 patent/WO2014059890A1/zh active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8098848B2 (en) * | 2006-05-19 | 2012-01-17 | Nuance Communications, Inc. | System for equalizing an acoustic signal |
CN101155438A (zh) * | 2006-09-26 | 2008-04-02 | 张秀丽 | 音频设备的频率响应自适应均衡方法 |
CN102447446A (zh) * | 2011-12-09 | 2012-05-09 | 苏州上声电子有限公司 | 基于振动元件运动状态反馈的扬声器频响均衡方法和装置 |
CN102883243A (zh) * | 2012-10-15 | 2013-01-16 | 苏州上声电子有限公司 | 在线迭代的声重放系统频响均衡方法和装置 |
CN102903367A (zh) * | 2012-10-15 | 2013-01-30 | 苏州上声电子有限公司 | 离线迭代的声重放系统频响均衡方法和装置 |
Non-Patent Citations (3)
Title |
---|
CAI, WEIJU ET AL.: "A New Variable Step Size LMS Adaptive Filtering Algorithm and Its Analysis.", COMPUTER & DIGITAL ENGINEERING., vol. 37, no. 8, 2009, pages 39 - 41 * |
GU, YUANTAO E ET AL.: "Novel variable step size NLMS algorithm.", J. TSINGHUA UNIV., vol. 42, no. 1, 2002, pages 15 - 18 * |
ZHANG, YABIN ET AL.: "Adaptive Equalization Simulation Based on LMS Algorithm.", JOURNAL OF ISHOU UNIVERSITY, vol. 27, no. 5, September 2006 (2006-09-01), pages 73 - 75 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115455745A (zh) * | 2022-11-09 | 2022-12-09 | 深圳飞骧科技股份有限公司 | 自适应频点采样的扫频方法、系统及相关设备 |
CN115455745B (zh) * | 2022-11-09 | 2023-04-07 | 深圳飞骧科技股份有限公司 | 自适应频点采样的扫频方法、系统及相关设备 |
Also Published As
Publication number | Publication date |
---|---|
CN102883243A (zh) | 2013-01-16 |
CN102883243B (zh) | 2015-03-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2014059890A1 (zh) | 在线迭代的声重放系统频响均衡方法和装置 | |
US8355510B2 (en) | Reduced latency low frequency equalization system | |
JP4402040B2 (ja) | リスニング領域内での、バスサウンドの音質を改善するためのイコライゼーションシステム | |
JP5241921B2 (ja) | 電子音響チャンネルの適応制御とイコライゼーションの方法 | |
CN106664473B (zh) | 信息处理装置、信息处理方法和程序 | |
JP6389259B2 (ja) | マイクロホンアレイを使用した残響音の抽出 | |
US8077880B2 (en) | Combined multirate-based and fir-based filtering technique for room acoustic equalization | |
Haneda et al. | Multiple-point equalization of room transfer functions by using common acoustical poles | |
CN108141691B (zh) | 自适应混响消除系统 | |
Schneider et al. | Adaptive listening room equalization using a scalable filtering structure in thewave domain | |
US20160149550A1 (en) | Method and apparatus to equalize acoustic response of a speaker system using multi-rate fir and all-pass iir filters | |
CN110289011B (zh) | 一种用于分布式无线声传感器网络的语音增强系统 | |
Talagala et al. | Efficient multi-channel adaptive room compensation for spatial soundfield reproduction using a modal decomposition | |
CN102447446A (zh) | 基于振动元件运动状态反馈的扬声器频响均衡方法和装置 | |
WO2022247494A1 (zh) | 音频信号补偿方法及装置、耳机、存储介质 | |
WO2014059897A1 (zh) | 离线迭代的声重放系统频响均衡方法和装置 | |
JP5762479B2 (ja) | 音声スイッチ装置、音声スイッチ方法、及びそのプログラム | |
JP3565846B2 (ja) | 適応音響システム及び音再生システム | |
US20080285768A1 (en) | Method and System for Modifying and Audio Signal, and Filter System for Modifying an Electrical Signal | |
WO2023051622A1 (zh) | 提升远场语音交互性能的方法和远场语音交互系统 | |
Jin | Adaptive reverberation cancelation for multizone soundfield reproduction using sparse methods | |
JP5937451B2 (ja) | エコー消去装置、エコー消去方法及びプログラム | |
WO2022209171A1 (ja) | 信号処理装置、信号処理方法およびプログラム | |
JP2006066988A (ja) | 収音方法、収音装置、収音プログラム、およびこれを記録した記録媒体 | |
Ignatov et al. | Semi-Automated Technique for Noisy Recording Enhancement Using an Independent Reference Recording |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 13847394 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 13847394 Country of ref document: EP Kind code of ref document: A1 |