WO2008087577A1 - Récepteur pour un signal audio multicanal et procédé de traitement d'un signal audio multicanal et dispositif de traitement de signal - Google Patents

Récepteur pour un signal audio multicanal et procédé de traitement d'un signal audio multicanal et dispositif de traitement de signal Download PDF

Info

Publication number
WO2008087577A1
WO2008087577A1 PCT/IB2008/050114 IB2008050114W WO2008087577A1 WO 2008087577 A1 WO2008087577 A1 WO 2008087577A1 IB 2008050114 W IB2008050114 W IB 2008050114W WO 2008087577 A1 WO2008087577 A1 WO 2008087577A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
frequency
sum
difference
predicted
Prior art date
Application number
PCT/IB2008/050114
Other languages
English (en)
Inventor
Andries P. Hekstra
Original Assignee
Nxp B.V.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nxp B.V. filed Critical Nxp B.V.
Publication of WO2008087577A1 publication Critical patent/WO2008087577A1/fr

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B1/00—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
    • H04B1/06—Receivers
    • H04B1/16—Circuits
    • H04B1/1646—Circuits adapted for the reception of stereophonic signals
    • H04B1/1661—Reduction of noise by manipulation of the baseband composite stereophonic signal or the decoded left and right channels
    • H04B1/1669—Reduction of noise by manipulation of the baseband composite stereophonic signal or the decoded left and right channels of the demodulated composite stereo signal
    • H04B1/1676—Reduction of noise by manipulation of the baseband composite stereophonic signal or the decoded left and right channels of the demodulated composite stereo signal of the sum or difference signal

Definitions

  • the present invention relates to a receiver for a multi- channel audio signal.
  • the present invention further relates to method for processing a multi-channel audio signal.
  • the present invention further relates to a signal processing device.
  • the demodulated FM-stereo signal comprises a mono audio signal (L+R), a pilot tone of 19 kHz and a stereo difference signal (L-R) modulated on 38 kHz sub carrier.
  • L+R mono audio signal
  • L-R stereo difference signal
  • the left and the right channel are reconstructed from the mono signal and the difference signal.
  • the received FM signal comprises white noise
  • the demodulated signal comprises a component that linearly increases with frequency.
  • the mono audio signal is present in the low frequency area (below 15 kHz) it contains a substantially lower noise level than the difference signal.
  • Known receivers therefore switch from stereo to mono operation in case that the input signal has too low a signal to noise ratio. In that case audible noise is low again.
  • the difference signal is reconstructed from the sum signal using a prediction factor, which is an estimation of the relation between the difference signal and the sum signal.
  • the prediction factor depends on a measure of similarity between the sum signal and the difference signal in a frequency range comprising that frequency. Use is made of the fact that the relation between the difference signal and the sum signal is relatively frequency independent. Even if the difference signal has a relatively strong noise component the said relation can be estimated relatively accurately by calculating an average over a range of frequencies. The average is calculated for a plurality of frequency ranges. At least one of the ranges is greater than 1. Signals are processed mainly in the frequency domain.
  • a frequency transformation is applied to transform the input signals into the frequency domain and after processing an inverse frequency transformation is applied to obtain time-domain signals again.
  • the frequency transformation is a Fourier transformation, e.g. implemented by a Fast Fourier transformation for example.
  • Alternative frequency transformation methods are available, such as a digital cosine transformation (DCT).
  • DCT digital cosine transformation
  • more complicated frequency transformations may be used that not only provide a decomposition of the signals into their frequency components, but for example into a combination of frequency components and time components. Otherwise stated, this results in one or more additional vectors of components of the sum signal and of the difference signal in the frequency domain.
  • Such a decomposition known as a time- frequency decomposition is described for example in "Binaural Cue coding - Part II: Schemes and Applications” by Christof Faller and Frank Baumgarte, in IEEE Transactions on Speech and Audio Processing, VoI 11, No. 6, November 2003.
  • Such time-frequency decompositions are used for example in audio coding, notably binaural audio coding, for which level, time (or phase) differences, applied to a mono-signal can produce a perceptually accurate illusion of stereo reproduction.
  • the application of such a decomposition to denoising of multi-channel audio signals is not shown however.
  • a constant range of frequencies may be used.
  • the method of claim 2 gives the best trade-off between a reduction of noise on the one hand and maintaining the information in the original difference signal on the other hand.
  • Predicted multi-channel signals may be obtained from the predicted difference signal in several ways. For example first a frequency domain version of the predicted multichannel signals may be obtained by a linear combination of the frequency domain version of the sum signal and the predicted difference signal. The time-domain versions of the predicted multi-channel signals may be obtained subsequently by applying an inverse Fourier transformation on the frequency domain versions of the predicted multi-channel signals.
  • the method of claim 3 is preferred however, as the received sum signal can be used, therewith partly avoiding consequences due to quantization errors in the frequency transformation of the sum signal. In addition, only a single inverse frequency transformation is required. A very practical implementation of the method that provides good quality results is claimed in claim 4.
  • the time windows may have a length in the order of 20-40 ms and comprise for example 1000-10000 samples.
  • the number of samples is a power of 2 to simplify the calculation.
  • An amount of samples in a time window substantially longer than 40 ms, e.g. 100 ms would necessitate complicated hardware to allow execution of a real time frequency transformation.
  • a time window substantially shorter than 20 ms, e.g. 5 ms would result in audible artifacts, in particular in the low frequency range.
  • the similarity measure is determined for a frequency range that comprises the frequencies for which the predicted difference component is computed.
  • Claim 7 describes a further preferred embodiment. By attenuating the prediction factor for those frequency ranges that have a relatively low signal to noise ratio a further improvement of the quality is achieved.
  • the method is preferably applied in a receiver as claimed in claim 10.
  • This can be a car radio receiver, an FM receiver embedded in a MP3 player, or in a mobile handset.
  • Claim 9 relates to a signal processing system for such a receiver.
  • the signal processor may be implemented in dedicated hardware, but may alternatively be implemented by a general purpose processor that performs the signal processing functions. If the signal processor is implemented in dedicated hardware, several functions may be implemented by the same module. For example the frequency transformation of the sum and the difference signal and the inverse frequency transformation of the difference signal may be carried out by one frequency transformation module.
  • Fig. 1 shows an FM-signal in the frequency domain
  • Fig. 2 schematically shows an FM-receiver
  • Fig. 3 shows a method according to the invention
  • Fig. 4 shows a signal processing device according to the invention
  • Fig. 5 shows a receiver according to the invention.
  • FM broadcasting is a technology invented to allow for a high-fidelity reproduction of audio signals.
  • S (L+R)/2
  • D (L-R)/2.
  • a mono receiver will use just the S signal.
  • the S signal is transmitted as baseband audio in the range 30 Hz to 15 kHz.
  • the D signal is amplitude-modulated onto a 38 kHz suppressed carrier to produce a double-sideband suppressed carrier (DSBSC) signal in the range 23 to 53 kHz.
  • DSBSC double-sideband suppressed carrier
  • the final multiplex signal from the stereo generator is the sum of the baseband audio (S), the pilot tone, and the DSBSC modulated subcarrier (D). This multiplex, along with any other subcarriers, modulates the FM transmitter.
  • Fig. 2 shows a block diagram of a typical FM receiver.
  • the FM input signal FM is first subjected to a limiter 10 in order to eliminate any amplitude modulation (AM) noise present in the signal.
  • the output of the limiter is a square wave with constant amplitude.
  • the square wave is then sent through the bandpass filter 20.
  • the band pass filter has a center frequency equal to the carrier frequency and a bandwidth equal to the bandwidth of the FM signal.
  • the bandpass filter 20 filters out the square wave harmonics and returns a constant-amplitude sinusoid.
  • the constant-amplitude FM signal is then differentiated in differentiator 30.
  • the instantaneous frequency is converted to an AM signal modulating the FM carrier function.
  • An envelope detector 40 extracts the amplitude, or envelope, of the input signal of interest. In this way the multiplex signal shown in Fig. 1 is retrieved.
  • a demultiplexer 50 derives a sum signal s(t) and a difference signal d(t) from
  • the difference signal which is present around the suppressed carrier at 38 kHz is significantly more affected than the mono signal in the range up to 15 kHz.
  • Some known receivers therefore automatically switch to mono audio reproduction if the level of noise is too high.
  • a method according to the present invention is presented below. This method enables stereo audio reproduction of reasonable quality even if a relatively high noise level is present in the input signal FM.
  • step Sl of the method for processing a multi-channel audio signal a sum signal s(t) representative for a sum of a first and a second audio signal, as well as a difference signal d(t) representative for a difference between the first and the second audio signal is received.
  • step S2 a frequency transformation of the sum and the difference signal is computed, resulting in a vector of components of the sum signal S(f) in the frequency domain and a vector of components of the difference signal D(f) in the frequency domain.
  • step S3 a covariance Cov(i) over the range i with size sz(i) between the components of the sum S(f) and the difference signal D(f) in the frequency domain is calculated.
  • Another measure for the covariance Cov may be applied which uses the difference signal D(f) for the lower frequency ranges and the norm
  • step S4 a variance Var(i) of the components of the sum signal S(f) over a range i in the frequency domain is calculated
  • step S5 a prediction factor C(i) is calculated for said frequency range i from the covariance Cov(i) and the variance Var(i), or from their alternatives Cov', Cov", and Var, or from Cov'" and Var'" for example.
  • c (0 Cov(0 Var ⁇ i)
  • the prediction factor C(i) is a product of a similarity measure p(i) and a scaling factor M(i) for frequency range i,
  • ⁇ d(i) is the standard deviation of the difference signal S(f) in the range i.
  • the prediction factor includes an attenuation An that depends on an estimation of the signal to noise ratio. This attenuation is not necessary, but provides for a further improvement of audio quality.
  • the factor An is preferably inversely proportional to a measure Edifr for the energy in the difference signal.
  • a moving average may be used for the estimation of this measure. For example a weighted geometrical mean Edifr between an energy E c measured in the current time-window and an energy E 'diff measured in the previous window.
  • E dlff E c a E l ⁇ f , wherein ⁇ is in the range 0.7 - 0.9, e.g. 0.8
  • the weighted mean may be applied recursively. I.e. in a subsequent window the value Ediff is substituted for E 'diff.
  • the energy of the difference signal in a window can be calculated as follows
  • the value of ⁇ is in the range of 1.2 to 1.6, e.g. 1.4. In this way a balanced estimation of the difference energy is obtained that is not influenced too much by peaks in the frequency spectrum. It has been found that spectrally narrow peaks mask the noise less than spectrally wider signals.
  • step S6 a predicted difference component D'(f) is calculated for each frequency within said frequency range i by multiplication of the prediction factor C(i) with the component of the sum signal S(f) for said frequency f,
  • step S7 control flow is returned to step S3 if the steps S3 to S6 have not been completed yet for all ranges i. If all ranges are traversed, control flow continues with step S8, wherein a predicted multi-channel audio signal is computed.
  • l'(t) s(t) + d'(t)
  • r'(t) s(t) - d'(t)
  • R(J) S(J) - D' (J) , wherein R(f) and L(f) are the frequency domain representation of the right and the left audio signal. These frequency domain representations can be subsequently transformed into their counterpart in the time-domain, by an inverse frequency transformation.
  • the size of the ranges i increases (preferably linearly) with the average frequency of the range.
  • the size sz(i) of range i is determined by:
  • the value of a which should be > 0, is preferably dependent on the number of samples n in a window, e.g.
  • ⁇ is in the range of 5 - 15, e.g. 10.
  • the method is preferably repeated for subsequent time-windows. For example the method is applied for each subsequent time-window of 20-40 ms.
  • the final predicted difference signal in the time domain signal is obtained by concatenating the predicted difference signals in the time domain.
  • the overlap-add method the frequency-transformation is preceded by multiplication with a first windowing function and the inverse frequency-transformation is succeeded by a second windowing function.
  • the first and the second windowing function may be the same.
  • the first windowing function is the square w 2 (t) of the windowing function w(t)
  • a block function may be used as the first windowing function and the function w 2 (t) may be used as the second windowing function.
  • the resulting data obtained after multiplication with the second window is added.
  • the overlap save-method may be used.
  • the first window is a block function having its edges coinciding with the selected time-window.
  • the second windowing function is a block function having half the width of the block function of the first window. This second windowing function is used to select the middle half of the samples of the processed result. Again subsequent time-windows overlap so that subsequent blocks of data selected by the second windowing function exactly complement each other.
  • a windowing function is not always necessary, for example when a digital cosine transformation is applied as the frequency transformation.
  • Fig. 4 shows a signal processing device for processing a multi-channel audio signal.
  • the device of Fig. 4 is coupled to an output of an FM-demodulator 60 that provides a sum signal s(t) representative for a sum of a first and a second audio signal and a difference signal d(t) representative for a difference between the first and the second audio signal.
  • the signal processing device comprises a facility 70 for applying a frequency transformation to the sum signal s(t) and a facility 80 for applying a frequency transformation to the difference signal d(t).
  • the result of the frequency transformation is a vector of components D(f) of the sum signal in the frequency domain and a vector of components S(f) of the difference signal in the frequency domain.
  • the device further has a facility 90 for calculating a covariance Cov(i) between the components of the sum S(f) and the difference signal D(f) in the frequency domain for a set of frequencies f in each of a plurality of frequency ranges i.
  • the device further has a facility 100 for calculating a variance Var(i) of the components of the sum signal S(f) in the frequency domain, for a set of frequencies f in each of a plurality of frequency ranges i.
  • a facility 110 calculates a prediction factor C(i) for each frequency range i from the covariance Cov(i) and the variance Var(i).
  • the prediction factor C(i) may depend on a value that indicates a level of audible noise An(i) within said frequency range.
  • the multiplication may also be carried out stepwise, e.g. by subsequent multiplication with components of the prediction factor, for example by a first multiplication with a similiarity measure, e.g. the cross correlation p(i) and a second multiplication with a scalingfactor M(i), or the other way around.
  • a facility 120 calculates a predicted difference component D'(f) of the frequency signal by multiplication of the prediction factor C(i) with the component of the sum signal S(f) for each frequency within each frequency range.
  • a facility is further included for computing a predicted multi-channel audio signal l'(t), r'(t).
  • the facility includes a facility 130 for calculating an inverse frequency transformation for calculating the time domain representation of the predicted difference component d'(t) and an adder 140 as well as a subtractor 150 for calculating linear combinations l'(t) and r'(t) of the signal d'(t) with the sum signal s(t).
  • Fig. 5 shows an embodiment of a receiver for a multi-channel audio signal.
  • the receiver comprises a module 210, for example as shown in Fig. 1, for receiving an FM signal FM and for demodulating the received signal into a sum signal s(t) and a difference signal d(t).
  • a signal processing device 220 according to the invention, for example as shown in Fig. 6, is included for processing the demodulated signals and providing reconstructed left and right audio signals l'(t) and r'(t).
  • a module 230, 240 is included for amplifying the predicted multi-channel audio signal l'(t), r'(t). Usually the high frequency audio signals are emphasized by the transmitter to improve the signal to noise ratio in the high frequency domain.
  • the receiver shown comprises de-emphasis modules 250, 260 to restore the power spectrum before reproduction.
  • the functions shown in Fig. 4 are all performed by software that is executed by a general purpose processor.
  • the system may implemented in hardware, or by a combination of software and hardware.
  • the system shown in Fig. 4 may be built from separate hardware components.
  • various functions may be combined.
  • the signal processing device of Fig. 4 may have a single frequency transformation facility that is used for transforming both signals d(t) and s(t) provided by the FM-demodulator and the reconstructed difference signal d'(f) on a time-shared basis.
  • n n _ f (SJf)-DXf)Y for f ⁇ n/2
  • the predicted complex audio signal in the time domain is then obtained by an inverse complex frequency transformation.
  • the facilities 90 may be reused to calculate the variance of the signal S(f) by providing the signal S(f) to each of the inputs of this facility 90.
  • the frequency domain is partitioned in a number of mutually disjoint frequency ranges. If a frequency range comprises more than one frequency, the prediction coefficient based on this frequency range is used for the calculation of the predicted difference signal components for each of the frequencies within the range.
  • moving frequency ranges may be used for calculating the prediction coefficient.
  • the prediction coefficient may be calculated for a range that if possible is symmetrically arranged around the frequency.
  • the calculation of the prediction coefficient may in addition include information from a time range than comprises the time interval for which the predicted difference signal component is calculated. For example, instead of using the prediction coefficient C(i) described above that is calculated for a time-interval tl-t2 an average value may be calculated from the prediction coefficients for the current time-interval tl-t2 a preceding time-interval tO-tl and a succeeding time-interval tl-t2.
  • the choice of the frequency ranges may depend on the distribution of the energy in the frequency spectrum of the signals.
  • the frequency ranges may have a size such that the energy content within each range exceeds a predetermined value, or such that the ranges substantially have the same energy content.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Stereo-Broadcasting Methods (AREA)

Abstract

La présente invention concerne un procédé de traitement d'un signal audio multicanal qui comprend les étapes consistant à : recevoir un signal de somme et un signal de différence représentatifs d'une somme d'un premier et d'un second signal audio et d'une différence entre le premier et le second signal audio, respectivement ; calculer une transformation fréquentielle du signal de somme et du signal de différence, ayant pour résultat un vecteur de composants du signal de somme dans le domaine fréquentiel et un vecteur de composants du signal de différence dans le domaine fréquentiel ; dans le domaine fréquentiel, calculer un composant prédiction du signal de différence par multiplication du composant du signal de somme pour ladite fréquence avec un facteur de prédiction qui dépend d'une mesure de similarité entre le signal de somme et le signal de différence dans une plage fréquentielle comprenant cette fréquence ; calculer un signal audio multicanal prédit, le calcul comprenant une transformation fréquentielle inverse.
PCT/IB2008/050114 2007-01-17 2008-01-14 Récepteur pour un signal audio multicanal et procédé de traitement d'un signal audio multicanal et dispositif de traitement de signal WO2008087577A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP07100663 2007-01-17
EP07100663.9 2007-01-17

Publications (1)

Publication Number Publication Date
WO2008087577A1 true WO2008087577A1 (fr) 2008-07-24

Family

ID=39315056

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/IB2008/050114 WO2008087577A1 (fr) 2007-01-17 2008-01-14 Récepteur pour un signal audio multicanal et procédé de traitement d'un signal audio multicanal et dispositif de traitement de signal

Country Status (1)

Country Link
WO (1) WO2008087577A1 (fr)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102201823A (zh) * 2010-03-25 2011-09-28 Nxp股份有限公司 多通道音频信号处理
CN102903363A (zh) * 2011-07-25 2013-01-30 哈曼贝克自动系统股份有限公司 立体声解码
EP2615739A1 (fr) 2012-01-16 2013-07-17 Nxp B.V. Processeur pour récepteur de signal FM et procédé de traitement

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5067157A (en) * 1989-02-03 1991-11-19 Pioneer Electronic Corporation Noise reduction apparatus in an FM stereo tuner
EP0955732A1 (fr) * 1998-04-29 1999-11-10 Sony International (Europe) GmbH Méthode et dispositif de démultiplexage d'un signal-multiplex-stéréo démodulé en fréquence
US20030039363A1 (en) * 2001-08-24 2003-02-27 Jens Wildhagen Noise reduction in a stereo receiver

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5067157A (en) * 1989-02-03 1991-11-19 Pioneer Electronic Corporation Noise reduction apparatus in an FM stereo tuner
EP0955732A1 (fr) * 1998-04-29 1999-11-10 Sony International (Europe) GmbH Méthode et dispositif de démultiplexage d'un signal-multiplex-stéréo démodulé en fréquence
US20030039363A1 (en) * 2001-08-24 2003-02-27 Jens Wildhagen Noise reduction in a stereo receiver

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102201823A (zh) * 2010-03-25 2011-09-28 Nxp股份有限公司 多通道音频信号处理
EP2369861A1 (fr) 2010-03-25 2011-09-28 Nxp B.V. Traitement de signal audio multi-canal
US20110235809A1 (en) * 2010-03-25 2011-09-29 Nxp B.V. Multi-channel audio signal processing
CN102201823B (zh) * 2010-03-25 2013-11-06 Nxp股份有限公司 多通道音频信号处理
US8638948B2 (en) 2010-03-25 2014-01-28 Nxp, B.V. Multi-channel audio signal processing
CN102903363A (zh) * 2011-07-25 2013-01-30 哈曼贝克自动系统股份有限公司 立体声解码
EP2552027A1 (fr) * 2011-07-25 2013-01-30 Harman Becker Automotive Systems GmbH Décodage stéréo
US9374117B2 (en) 2011-07-25 2016-06-21 Harmon Becker Automotive Systems Gmbh Stereo decoding system
CN105743526A (zh) * 2011-07-25 2016-07-06 哈曼贝克自动系统股份有限公司 立体声解码
EP2615739A1 (fr) 2012-01-16 2013-07-17 Nxp B.V. Processeur pour récepteur de signal FM et procédé de traitement
US9172479B2 (en) 2012-01-16 2015-10-27 Nxp, B.V. Processor for an FM signal receiver and processing method

Similar Documents

Publication Publication Date Title
CN102113315B (zh) 用于处理音频信号的方法和装置
JP5101579B2 (ja) 空間的オーディオのパラメータ表示
RU2584009C2 (ru) Обнаружение высокого качества в стереофонических радиосигналах с частотной модуляцией
KR101137359B1 (ko) 다이알로그 증폭 기술
EP2612322B1 (fr) Procédé et appareil de décodage d'un signal audio multicanal
RU2526745C2 (ru) Низведение параметров последовательности битов sbr
US8090122B2 (en) Audio mixing using magnitude equalization
US20100086136A1 (en) Spatial disassembly processor
WO2009046225A2 (fr) Procédé de corrélation pour l'extraction d'ambiance de signaux audio à deux canaux
CN105284133B (zh) 基于信号下混比进行中心信号缩放和立体声增强的设备和方法
WO2015003900A1 (fr) Procédé et appareil pour générer, à partir d'une représentation de domaine coefficient de signaux ambiophoniques d'ordre supérieur, une représentation de domaine mixte spatial/coefficient desdits signaux ambiophoniques d'ordre supérieur
US8638948B2 (en) Multi-channel audio signal processing
WO2008087577A1 (fr) Récepteur pour un signal audio multicanal et procédé de traitement d'un signal audio multicanal et dispositif de traitement de signal
TW201532035A (zh) 預測式ｆｍ立體聲無線電雜訊降低
US9311925B2 (en) Method, apparatus and computer program for processing multi-channel signals
US20160344902A1 (en) Streaming reproduction device, audio reproduction device, and audio reproduction method
US10003422B2 (en) Method for processing an FM stereo signal
CN107358961B (zh) 多声道信号的编码方法和编码器
Cole et al. Frequency offset correction for HF radio speech reception
EP2709101A1 (fr) Système et procédé de traitement audio numérique

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: 08702414

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: 08702414

Country of ref document: EP

Kind code of ref document: A1