WO2005120133A1

WO2005120133A1 - Apparatus and method of reproducing wide stereo sound

Info

Publication number: WO2005120133A1
Application number: PCT/KR2005/001559
Authority: WO
Inventors: Sun-Min Kim
Original assignee: Samsung Electronics Co., Ltd.
Priority date: 2004-06-04
Filing date: 2005-05-27
Publication date: 2005-12-15
Also published as: JP2008502200A; EP1752017A1; EP1752017A4

Abstract

An apparatus and a method of reproducing a wide stereo sound by widening a stereo sound output by an audio reproducing apparatus using only two closely disposed channel speakers include a widening filtering operation and a direct filtering operation. In the widening filtering operation, virtual sound sources for arbitary locations are formed from a stereo-channel audio signal using head related transfer functions measured at predetermined locations, and crosstalk is cancelled from the virtual sound sources using filter coefficients in which the head related transfer functions are reflected. In the direct filtering operation, signal characteristics of the stereo-channel audio signal are adjusted based on the crosstalk-cancelled virtual sound sources.

Description

Description APPARATUS AND METHOD OF REPRODUCING WIDE STEREO SOUND Technical Field

[1] The present general inventive concept relates to an audio reproduction system, and more particularly, to a method and an apparatus to reproduce a wide stereo sound by widening a stereo sound output by an audio reproducing apparatus using only speakers of two channels that are disposed close to each other. Background Art

[2] Since televisions generally include speakers of two channels attached to either the right and the left or the bottom of a main body, a hearing angle is narrow. Hence, a stereo effect generated by DVD/CD reproducers or a television broadcast is reduced, and stereo sounds are heard like mono sounds. In particular, a narrow stereo sound stage reduces the sound quality of a movie and can cause movie viewers to purchase extra speaker systems.

[3] Conventional stereo enhancement systems enhance stereo sounds in front of a listener using only two speakers.

[4] A conventional stereo enhancement system is disclosed in U.S. Patent No. 6,597,791 (filed on December 15, 1998), entitled Audio Enhancement System.'

[5] Referring to U.S. Patent No. 6,597, 791, the conventional stereo enhancement system processes a difference signal generated from left and right input signals to create a stereo sound. The difference signal is processed through equalization characterized by amplification of auditory frequencies of high and low bands. The processed difference signal is combined with a sum signal, generated from the left and right input signals, and the original left and right input signals. Disclosure of Invention Technical Problem

[6] However, most conventional stereo enhancement systems have difficulties in designing a crosstalk cancellation filter, so they either use a sum of right and left channels of a stereo sound and a difference between the right and left channels or adjust a phase of and an amplitude of the stereo sound, instead of using a head related transfer function (HRTF). The non-use of HRTFs reduces the amount of calculation required by the conventional stereo enhancement systems, so the conventional stereo enhancement systems can be easily implemented. However, the conventional stereo enhancement systems do not have excellent performances because they are designed without consideration of a head and an auricle of a human being. Technical Solution

[7] The present general inventive concept provides a method of reproducing a wide stereo sound by widening a stereo sound stage output by an audio reproducing apparatus using only speakers of two channels that are disposed close to each other.

[8] The present general inventive concept also provides an apparatus to reproduce a wide stereo sound according to the above-described method Advantageous Effects

[9] In a wide stereo reproducing apparatus and method according to the present general inventive concept, a widening filter is obtained by convolving a binaural synthesis portion with a crosstalk canceller to thereby reduce calculations. Also, sounds are output not only through virtual speakers using HRTFs but also through actual speakers. A panorama filter is designed to be a matrix in which the widening filter coefficients for the virtual speakers and direct filter coefficients for the actual speakers are convolved. Each of the filters is designed to have an optimal performance, and the optimal performance is maintained through various hearing tests. Due to the use of frequency sampling, each of the filter coefficients has an optimal performance and minimizes the amount of calculation. Thus, when the wide stereo reproducing apparatus and method according to the present general inventive concept are applied to products having two closely arranged speakers, such as, TVs, PCs, Note PCs, PDAs, cellular phones, and the like, a stereo sound stage is widened, so listeners can feel an enhanced stereo sound effect without need to purchasing extra speaker sets. Description of Drawings

[10] FIG. 1 is a block diagram illustrating an apparatus to reproduce a wide stereo sound, according to an embodiment of the present general inventive concept;

[11] FIG. 2 is a flowchart illustrating a method of implementing the apparatus of FIG. 1;

[12] FIG. 3 is a detailed block diagram illustrating binaural synthesis portions of the apparatus of FIG. 1 ;

[13] FIG. 4 is a detailed block diagram illustrating a crosstalk canceller of the apparatus of FIG. 1;

[14] FIG. 5 is a block diagram illustrating a matrix relationship between a pair of direct filters and a widening filter of the apparatus of FIG. 1 ;

[15] FIG. 6 is a conceptual diagram illustrating a panorama filter of the apparatus of FIG. 1;

[16] FIG. 7 is a block diagram illustrating a production of a wide stereo sound from a mono sound according to an embodiment of the present general inventive concept; and

[17] FIG. 8 is a block diagram illustrating a production of an adaptive wide stereo sound according to an embodiment of the present general inventive concept. Best Mode

[18] The foregoing and/or other aspects and advantages of the present general inventive concept may be achieved by providing a method of reproducing a stereo sound in an audio reproducing apparatus, the method including a widening filtering operation and a direct filtering operation. In the widening filtering operation, virtual sound sources corresponding to arbitrary locations are formed from a stereo-channel audio signal using head related transfer functions measured at predetermined locations, and crosstalk is cancelled from the virtual sound sources using filter coefficients in which the head related transfer functions are reflected. In the direct filtering operation, signal characteristics of the stereo-channel audio signal are adjusted based on the crosstalk- cancelled virtual sound sources.

[19] The foregoing and/or other aspects and advantages of the present general inventive concept may also be achieved by providing a method of reproducing a stereo sound in an audio reproducing apparatus, the method comprising a stereo-channel audio signal receiving operation of receiving a stereo-channel audio signal, and a panorama filtering operation. In the panorama filtering operation, virtual sound sources are formed from the stereo-channel audio signal, crosstalk is cancelled from the virtual sound sources, and signal characteristics of the input stereo-channel audio signal are adjusted based on the crosstalk-cancelled virtual sound sources. The adjusting of the signal characteristics of the input stereo-channel audio signal may be expressed as the following equation:

[20] yL = P 11 (z)L + P 12 (z)R [21] y ^ R = P 21 (z)L + P 22 (z)R, [22] wherein L and R denote left and right input signals of two channels, respectively, and y and y denote left and right output signals, respectively . Filter coefficients P (z), P (z), P (z), and P (z) may be calculated using the following equation:

[23] ^Put 7_n (z) + D_L(z) W_u (z) ^■

[24] wherein W(z) is expressed in the following equation:

[25]

[26] 1. and D(z) denotes a diagonal matrix comprising filter coefficients (D (z), D (z)) having a delay time and an amplitude of the stereo-channel audio signal.

[27] The foregoing and/or other aspects and advantages of the present general inventive concept may also be achieved by providing an apparatus to reproduce a stereo sound, the apparatus including a binaural synthesis portion, a crosstalk canceller, and direct filters. The binaural synthesis portion forms virtual sound sources corresponding to arbitrary locations from a stereo-channel audio signal using head related transfer functions measured at predetermined locations. The crosstalk canceller cancels crosstalk from the virtual sound sources formed by the binaural synthesis portion, using filter coefficients based on information about angles at which actual speakers are disposed. The direct filters adjust a signal amplitude of and a time delay of the stereo- channel audio signal based on the crosstalk-cancelled virtual sound sources using filter coefficients of the direct filters. Mode for Invention

[28] FIG. 1 is a block diagram illustrating an apparatus to reproduce a wide stereo sound, according to an embodiment of the present general inventive concept. Referring to FIG. 1, the apparatus includes a widening filter 120 and left and right direct filters 140 and 150. The widening filter 120 is formed by convolving left and right binaural synthesis portions 122 and 124 and a crosstalk canceller 128 together. A panorama filter 100 is formed by convolving the widening filter 120 with the left and right direct filters 140 and 150.

[29] The left and right binaural synthesis portions 122 and 124 produce virtual sound sources from a 2-channel audio signal based on head related transfer functions (HRTFs) measured at predetermined locations (angles) with respect to a sound source. In other words, the left and right binaural synthesis portions 122 and 124 render virtual speakers 182 and 192 symmetrically disposed in front of a listener, using the HRTFs. A left-channel audio signal of the 2-channel audio signal is convolved with HRTFs measured at - 30 degrees. Likewise, a right-channel audio signal of the 2-channel audio signal is convolved with HRTFs measured at +30 degrees. Hence, an audio signal convolved with the HRTF for the left ear at - 30 degrees and an audio signal convolved with the HRTF for the left ear at +30 degrees are summed to form a left virtual audio signal corresponding to a left virtual speaker 182. An audio signal convolved with the HRTF for the right ear at - 30 degrees and an audio signal convolved with the HRTF for the right ear at +30 degrees are summed to form a right virtual audio signal corresponding to a right virtual speaker 192.

[30] The crosstalk canceller 128 cancels crosstalk between the left and right virtual audio signals formed by the left and right binaural synthesis portions 122 and 124, based on filter coefficients in which the HRTFs are reflected. In other words, the crosstalk canceller 128 cancels the crosstalk between the left and right virtual audio signals so that the listener cannot hear the left virtual audio signal corresponding to the left virtual speaker 182 through the right ear and cannot hear the right virtual audio signal corresponding to the right virtual speaker 192 through the left ear.

[31] The left and right direct filters 140 and 150 adjust a level of and an output timing of the 2-channel audio signal with respect to the left and right virtual audio signals of which the crosstalk has been canceled by the crosstalk canceller 128. The left and right direct filters 140 and 150 can filter an input stereo sound and adjust an output timing of and a signal level of a sound to be output through actual speakers 180 and 190 with respect to a sound (left and right virtual audio signals) corresponding to the virtual speakers 182 and 192 to thereby produce a natural sound.

[32] The 2-channel audio signal filtered by the left and right direct filters 140 and 150 and the left and right virtual audio signals filtered by the widening filter 120 are summed and output to left and right actual speakers 180 and 190. Thus, the left and right actual speakers 180 and 190 output the 2-channel audio signal adjusted by the left and right direct filters 140 and 150 and the left and right virtual audio signals so that the listener hears the adjusted 2 channel audio signal from the left and right actual speakers 180 and 190, and the listener hears the left and right virtual audio signals from the left and right virtual speakers 182 and 192 although outputs (left and right audio signals of the 2-channel audio signal) of the left and right direct filters 140 and 150 and the left and right virtual audio signals of the widening filter 120 are output through the left and right actual speakers 180 and 190, respectively.

[33] FIG. 2 is a flowchart illustrating a method of implementing the apparatus of FIG. 1. An acoustic transfer function between a speaker and an eardrum is referred to as an HRTF. The HRTF contains information representing characteristics of a space into which a sound is transferred, including a difference between timings when sound wave signals reach the right and left ears, a difference between levels of the sound wave signals for the right and left ears, and shapes of the right and left pinnas. Particularly, the HRTF can include information about the pinnas that critically affect localizations of upper and lower sound images. The information about the pinnas can be obtained through measurements because modeling the pinnas is not easy.

[34] Referring to FIG. 2, at operation 212, angles at which the virtual speakers 182 and 192 are disposed are selected. At operation 216, the virtual speakers 182 and 192 are disposed based on binaural synthesis. The virtual sound sources can be formed at arbitrary locations by the use of an HRTF database measured at predetermined locations (angles) with respect to the speakers 180 and 190 and/or the virtual speakers 182 and 192. For example, if an HRTF measured at 30 degrees and an actual sound source are convolved, a sense of a virtual sound source at 30 degrees can be obtained. 2N virtual speakers are symmetrically disposed in front of a listener to widen a stereo sound stage. Right- and left-channel signals of a stereo sound pass through N virtual speakers located on the right side of the listener and N virtual speakers located on the left side of the listener, respectively.

[35] As illustrated in FIG. 3, a total of four HRTFs, including the two HRTFs between the left virtual speaker 182 and each of the right and left ears of the listener and the two HRTFs between the right virtual speaker 192 and each of the right and left ears, can be required to arrange the two virtual speakers 182 and 192. Accordingly, 4N HRTFs are required to arrange 2N virtual speakers. Since the 4N HRTFs can be represented as a sum of 2x2 square matrixes, when the sum is calculated using Equation 1, only a total of 4 HRTFs are required. Thus, an amount of calculation is drastically reduced. Equation 1 is:

[36] JV JV ^'L_L(z) R_L(z) ∑ iϋ Cz) ∑ _ΔCz⁾ 2-1 2-1 (1) L (z) R (z _ ∑^R _* Z -l

[37] wherein L Li (z) denotes an HRTF between an i-th left virtual speaker and the left ear, R Li (z) denotes an HRTF between an i-th right virtual speaker and the left ear, L Ri (z) denotes an HRTF between the i-th left virtual speaker and the right ear, and R Ri (z) denotes an HRTF between the i-th right virtual speaker and the right ear. [38] At operation 214, information regarding angles at which the actual speakers 180 and 190 are disposed is determined. At operation 218, the crosstalk canceller 128 based on an infinite impulse response (IIR) filter having an optimized performance is designed according to the information regarding the angles at which the actual speakers 180 and 190 are disposed. The crosstalk canceller 128 is used to prevent a stereo sound effect from being degraded due to generation of crosstalk between the two actual speakers 180 and 190 and the two ears of the listener upon sound reproduction through only the two actual speakers 180 and 190. FIG. 4 is a detailed block diagram of the crosstalk canceller 128. Referring to FIG. 4, d(z) denotes a binaural-synthesized signal, u(z) denotes an output of a speaker, and e(z) denotes an error to be minimized. Reference character H(z) denotes a transfer function matrix (e.g., a 2 x 2 square matrix) between two speakers and two ears of a listener, and reference character C(z) denotes a crosstalk-cancellation matrix designed to be inverse to the transfer function matrix H(z). Reference numeral A(z) denotes a pure delay filter matrix to satisfy causality. Since the transfer function matrix H(z) can have a shape of a finite impulse response (FIR) filter, the crosstalk-cancellation matrix C(z) can have a shape of an IIR filter because the crosstalk-cancellation matrix C(z) is inverse to the transfer function matrix H(z). However, because of stability, the crosstalk-cancellation matrix C(z) can be approximated to an FIR filter. In this case, despite the fact that the crosstalk cancellation matrix C(z) can be well approximated to a FIR filter of a high order, the crosstalk cancellation matrix C(z) can be approximated to an FIR filter of a low order, as well, because of hardware problems. Hence, obtaining an exact crosstalk cancellation matrix C(z) is difficult. The wide stereo sound reproducing apparatus of FIG. 1 can include a portion to convert an IIR filter into an FIR filter and optimize the order of the filter, such that an optimized IIR filter can be applied to a crosstalk canceller. The crosstalk cancellation matrix C(z) designed based on IIR filter coefficients is divided into a stable portion and an unstable portion. The stable portion is formed of the IIR filter, and the unstable portion is formed of the FIR filter. The two portions are convolved to obtain a single stable IIR filter.

[39] The number of and the locations of the virtual speakers 182 and 192 that affect binaural synthesis are predetermined, and the locations of the actual speakers 180 and 190 that affect the crosstalk canceller 128 are also predetermined. Hence, at operations 220 and 222, the binaural synthesis and the crosstalk canceller 128 are convolved to design the widening filter 120 based on the IIR filter. If 2N virtual speakers are arranged, a binaural synthesis is a 2x2 square matrix, and the crosstalk cancellation matrix C(z) is also a 2x2 square matrix. Hence, the widening filter is a 2x2 square matrix corresponding to a product of the two 2x2 square matrixes. The widening filter is obtained by Equation 2:

[40]

w₂₂ z) C_2l(z) C₂₂ (z) (z) R_& (z) ^■ )

[41] wherein W(z) denotes a widening filter matrix, C(z) denotes the crosstalk cancellation matrix, L (z) denotes the HRTF between the left virtual speaker 182 and the left ear, R (z) denotes the HRTF between the right virtual speaker 192 and the left ear, L (z) denotes the HRTF between the left virtual speaker 182 and the right ear, and R R R (z) denotes the HRTF between the right virtual speaker 192 and the right ear. [42] However, since the crosstalk canceller 128 is optimized based on the IIR filter, the order of the widening filter 120 can be increased like the crosstalk canceller filter 128. Thus, there can be difficulty in implementing the widening filter 120 in real time. Accordingly, at operation 224, the widening filter 120 converts the IIR filter into the FIR filter using frequency sampling to minimize the order of the widening filter. At this time, a frequency interval in a frequency band is adjusted using the frequency sampling to thereby adjust the order of the FIR filter. A minimum filter order that does not degrade a performance of a filter is determined through a hearing test.

[43] Thereafter, at operation 226, it is determined whether a performance test of the widening filter 120 through hearing experiments has been completed. When the performance test is completed, the direct filters 140 and 150 to correct a time delay and an output level difference between the actual speakers 180 and 190 and the virtual speakers 182 and 192 are designed, at operation 228. In other words, when the stereo sound passes through the widening filter 120 and is then reproduced through only the two actual speakers 180 and 190, the stereo sound seems to be reproduced through virtual speakers 182 and 192 arranged widely in front of the listener. In this case, although the stereo sound is widened by the widely arranged virtual speakers 182 and 192, the sound seems empty at the center of the front side of the listener where no virtual speakers 182 and 192 are disposed. Hence, the listener hears an unnatural sound having a deteriorated tone. To solve this problem, the direct filters 140 and 150 are designed so that the actual speakers 180 and 190 can also output sounds. The direct filters 140 and 150 adjust the sizes of outputs of the actual and virtual speakers 180, 190, 182 and 192 and a time delay between the actual and virtual speakers 180, 190, 182, and 192. The time delay by the direct filters 140 and 150 is matched with a predesigned time delay by the widening filter 120 to prevent a deterioration of the tone of the sound. The direct filters 140 and 150 determine a ratio of output levels of the actual speakers 180 and 190 to output levels of the virtual speakers 182 and 192. Thus, the direct filters can adjust a degree to which the stereo sound is divided. If the amplitude of each of the direct filters 140 and 150 is close to 0, the sound is reproduced through only the virtual speakers, and accordingly the sound from the center of the front side of the listener is empty although a stereo sound stage is widened. If the amplitude of each of the direct filters 140 and 150 is extremely large, the sound is reproduced through only the actual speakers 180 and 190, and accordingly a wide stereo effect is not obtained. Thus, the amplitudes of the direct filters 140 and 150 must be determined through a number of hearing tests. FIG. 5 is a block diagram illustrating a relationship between a matrix D(z) of each of the direct filters 140 and 150 and the matrix W(z) of the widening filter 120. The widening filter 120 forms the left and right virtual audio signals from the input stereo sound and outputs the left and right virtual audio signals corresponding to the virtual speakers 182 and 192. The direct filters 140 and 150 adjust signal characteristics of the input stereo sound based on the left and right virtual audio signals and outputs an adjusted input stereo sound to the actual speakers 180 and 190.

[44] At operation 232, a panorama filter 100 is designed by convolving the widening filter 120 and the direct filters 140 and 150. In other words, a parameter filter matrix P(z), which is a single filter, is obtained by adding the widening filter matrix W(z) and the direct filter matrix D(z). The panorama filter matrix P(z) is defined as in Equation 3: [45] P(z) = W(z) + D(z) ...(3)

[46] Each element of the matrix P(z) is calculated using Equation 4 :

[47] _n (z) P_u (z W_u(z + D_L (z) W_u (z (4) ₃₁ (z) R_Λ2 (z)_ ₂₁ (z) W _l (_.z) + D_R (z)_

[48] 1. wherein each element of the matrixes P(z) and W(z) is an FIR filter coefficient, and D(z) denotes a diagonal matrix comprising filter coefficients (D (z), D R (z)) having a pure delay time and a pure size.

[49] FIG. 6 illustrates the panorama filter 100 to reproduce the wide stereo sound. Referring to FIG. 6, since the stereo sound is a 2 x 2 vector, when the stereo sound passes through the panorama filter 100 in the shape of a 2 x 2 square matrix, a 2-channel widened stereo sound is output. The amplitude of a signal not yet passed through the panorama filter 100 and a signal passed through the panorama filter 100 can be adjusted through various hearing tests to obtain the greatest sound quality when the wide stereo sound is played. The values of the final output signals are obtained using Equation 5:

[50] y = P (z)L + P (z)R

[51] y R = p" 21(z)L + P 22 z)R ...(5)

[52] 1. wherein L and R denote left and right input signals of two channels, respectively, and y L, and y R denote left and right output signals of two channels, respectively.

[53] At operation 234, it is determined whether a performance test for the panorama filter through the hearing experiments has been completed. When the performance test is completed, the wide stereo sound is reproduced, in operation 236. Consequently, as illustrated in FIG. 6, a listener can hear a wide stereo sound through the actual speakers 180 and 190 and the virtual speakers 182 and 192.

[54] FIG. 7 is a block diagram of an apparatus to reproduce a wide stereo sound from a mono sound, according to an embodiment of the present general inventive concept.

[55] TV broadcasting stations generally output mono-sounds. The panorama filter matrix P(z), of FIG. 6 has a symmetrical structure as shown in Equation 4. Hence, when the mono-sound passes through the panorama filter matrix P(z), identical signals are output to the actual speakers 180 and 190. In other words, when the mono-sound is input to the panorama filter 100 of FIG. 6, a stereo sound effect is not generated. Referring to FIG. 7, the mono audio signal input through a single channel is converted into a 2-channel audio signal while passing through a phase inverter 710, which inverts a phase of the input mono signal by 180 degrees. The input mono audio signal and a mono audio signal having a 180 ° -converted phase are input to a panorama filter 100, which is pre-designed with an optimal filter. The stereo sound produced from the mono sound can be expressed as in Equation 6:

[56] L = M, R= - M ...(6)

[57] wherein L denotes a left channel, R denotes a right channel, and M denotes the mono sound.

[58] FIG. 8 is a block diagram of a system to produce an adaptive wide stereo sound, according to an embodiment of the present general inventive concept.

[59] When the wide stereo technology of FIG. 1 is used, the listener feels an optimal performance when the user is at a sweet spot. Since the location of the listener is generally not restricted, an optimal wide stereo performance should be obtained no matter where the listener is located. Thus, in the system of FIG. 8, a location of the listener is ascertained in real time, and the wide stereo sound is reproduced using filter coefficients pre-designed according to the ascertained location of the listener.

[601 Referring ^σ to FIG. 8, first, coefficients P 11 , P 12 , P 21 , and P 22 of the o ^rptimized panorama filter 100 corresponding to various locations of a listener are calculated. The panorama filter coefficients are stored in a filter coefficient table 820, which is a lookup table. A location ascertaining unit 810 ascertains a location of the listener using an iris recognition technology. The location ascertaining unit 810 is not limited to using the iris recognition technology, but may variously determine the location of the user. A controller 830 reads the filter coefficients P , P , P , and P corresponding to 11 12 21 22 the listener's location ascertained by the location ascertaining unit 810 from the filter coefficient table 820 and outputs the filter coefficients P 11 , P 12 , P 21 , and P 22 to the panorama filter 100. The panorama filter 100 generates the stereo sound corresponding to the input 2-channel audio signal using the received filter coefficients P 11 , P 12 , P 21 , and P . Consequently, the system of FIG. 8 can provide the stereo sound effect adaptive to each location of the listener. Industrial Applicability

[61] The general inventive concept can also be embodied as computer readable codes on a computer readable recording medium. The computer readable recording medium can be any data storage device that can store data which can be thereafter read by a computer system. Examples of the computer readable recording medium include readonly memory (ROM), random-access memory (RAM), CD-ROMs, magnetic tapes, floppy disks, optical data storage devices, and carrier waves (such as data transmission through the Internet). The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

Claims

[1] 1. A method of reproducing a stereo sound in an audio reproducing apparatus, the method comprising: forming virtual sound sources corresponding to arbitrary locations from a stereo- channel audio signal using head related transfer functions measured at predetermined locations, and canceling crosstalk between the virtual sound sources using filter coefficients in which the head related transfer functions are reflected in a widening filtering operation; and adjusting signal characteristics of the stereo-channel audio signal based on the crosstalk-cancelled virtual sound sources.

2. The method of claim 1, wherein the forming of the virtual sound sources comprises convolving the head related transfer functions with the stereo-channel audio signal to form the virtual sound sources in a binaural synthesis operation, and the canceling of the crosstalk comprises canceling the crosstalk in a crosstalk canceling operation.

3. The method of claim 2, wherein the convolving the head related transfer functions with the stereo channel audio signal comprises forming the virtual sound sources using coefficients calculated using the following equation:

where L Li (z) denotes a head related transfer function between an i-th left virtual speaker and a left ear of a listener, R Li (z) denotes a head related transfer function between an i-th right virtual speaker and the left ear, L Ri (z) denotes a head related transfer function between the i-th left virtual speaker and a right ear of the listener, and R Ri (z) denotes a head related transfer function between the i-th right virtual speaker and the right ear.

4. The method of claim 2, wherein a matrix of the filter coefficients for the crosstalk cancellation operation is inverse to a matrix of the head related transfer functions between two virtual speakers and right and left ears of a listener.

5. The method of claim 1, wherein the forming of the virtual sound sources comprise forming the virtual sound sources using a widening filter having coefficients calculated using the following equation: W_u (z) W_u (z) C_n (z) C_u (z) L_L(z) R_L(z) W_2l(z) W₂₂ (z)_ C_2l(z) C₂₂ (Z)_ L (z) R_£ z\

where W(z) denotes a widening filter coefficient, C(z) denotes a crosstalk canceller coefficient, L (z) denotes an HRTF between a left virtual speaker and the left ear, R (z) denotes an HRTF between a right virtual speaker and the left ear, L (z) denotes an HRTF between the left virtual speaker and the right ear, R and R (z) denotes an HRTF between the right virtual speaker and the right ear. R

6. The method of claim 1, wherein the forming of the virtual sound sources comprises converting high-order infinite impulse response filter coefficients into low-order finite impulse response filter coefficients using frequency sampling.

7. The method of claim 1, wherein, the signal characteristics comprise a signal amplitude and a time delay.

8. The method of claim 1, further comprising: forming a 2-channel stereo sound from an input mono sound by converting a phase of the input mono sound by 180 degrees.

9. A method of reproducing a stereo sound in an audio reproducing apparatus, the method comprising: receiving a stereo-channel audio signal; and forming virtual sound sources from the stereo-channel audio signal, canceling crosstalk from the virtual sound sources, and adjusting signal characteristics of the input stereo-channel audio signal based on the crosstalk-cancelled virtual sound sources in a panorama filter operation, wherein: the virtual sound sources are expressed as the following equation: y = P (z)L + P (z)R

" L 11 12 y ^ R = P 21 (z)L + P 22 (z)R where L and R denote left and right input signals of two channels, respectively, and y and y denote left and right output signals, respectively , and P (z), P 12 (z),

P 21 (z), and P 22 (z), and the filter coefficients are calculated using σ the following σ equation: _n(z) P_u P₂₁ (z) P₂₂

where W(z) is expressed in the following equation: _n (z) W_u (z) ^'C_u (z) C₁₂ (z W_Ά {Z) W₂₂ (Z) C_Ά (Z) C₂₂ (Z)

and D(z) denotes a diagonal matrix comprising filter coefficients (D (z), D (z)) L, R having a delay time and an amplitude of the stereo-channel audio signal.

10. The method of claim 9, wherein orders of the filter coefficients are adjusted by controlling a frequency interval in a frequency band.

11. The method of claim 9, further comprising: calculating the filter coefficients for the panorama filtering operation according to a location of a listener; detecting a location of the listener; reading filter coefficients for the panorama filtering operation corresponding to a detected location of the listener; and producing a stereo sound from the stereo-channel audio signal using the read-out filter coefficients.

12. An apparatus to reproduce a stereo sound, comprising: a binaural synthesis portion to form virtual sound sources corresponding to arbitrary locations from a stereo-channel audio signal using head related transfer functions (HTRF) measured at predetermined locations; a crosstalk canceller to cancel crosstalk from the virtual sound sources formed by the binaural synthesis portion, using filter coefficients based on information about angles at which actual speakers are disposed; and direct filters to adjust an amplitude of and a time delay of the stereo-channel audio signal based on the crosstalk-cancelled virtual sound sources.

13. The apparatus of claim 12, wherein the binaural synthesis portion and the crosstalk canceller act as a widening filter having a widening filter coefficient matrix formed by convolving an HRTF coefficient matrix of the binaural synthesis portion with a filter coefficient matrix of the crosstalk canceller, and calculated using the following equation:

^■W_u (z) W_u (z) C_n (z) C_l2 (z) L_L(z) R_L(z) W_2l z) W₂₂ (z) C_2l z) C₂₂ (z) L_M(z) R_& z)

wherein W(z) denotes a widening filter coefficient, C(z) denotes a crosstalk canceller coefficient, L (z) denotes an HRTF between a left virtual speaker and a left ear of a listener, R (z) denotes an HRTF between a right virtual speaker and the left ear of the listener, L R (z) denotes an HRTF between the left virtual speaker and a right ear of the listener, and R R (z) denotes an HRTF between the right virtual speaker and the right ear of the listener.

14. The apparatus of claim 13, wherein the binaural synthesis portion, the crosstalk canceller, and the direct filters act as a panorama filter having a panorama filter coefficient matrix formed by convolving the widening filter coefficient matrix with coefficients of the direct filters, and calculated using the following equation:

^~P_u (z) P_u (z) W_u (z)+ D_L(z) W_u (z) P_2l(z P₂₂ (z)_ W _l (z) FF₂₁ (z) + D_k (z)_

wherein D(z) denotes a diagonal matrix comprising direct filter coefficients (D (z), D R (z)) having only a delay time and an amplitude of the stereo-channel audio signal.

15. The apparatus of claim 14, further comprising: a filter coefficient table to store panorama filter coefficients according to a location of the listener; a location ascertaining unit to ascertain the location of the listener; and a controller to read the panorama filter coefficients corresponding to the location of the listener ascertained by the location ascertaining unit from the filter coefficient table and to output the corresponding panorama filter coefficients.

16. The apparatus of claim 12, further comprising a phase inverter to invert a phase of a mono sound to convert the mono sound into the stereo sound.

17. An apparatus to reproduce an input sound signal, comprising: first and second direct filters to respectively filter first and second channel signals of an input sound signal to adjust characteristics of the first and second channel signals; a widening filter comprising: first and second binaural portions to form first and second virtual signals according to the input first and second channel signals and head related transfer functions (HTRF), and a crosstalk canceller to cancel crosstalk between the first and second virtual signals according to the head related transfer functions; and a first output terminal to output the filtered first channel signal and the first virtual signal; and a second output terminal to output the filtered second signal and the second virtual signal.

18. The apparatus of claim 17, wherein the first and second direct filters filter the first and second channel signals according to the first and second virtual signals.

19. The apparatus of claim 17, wherein the characteristics of the first and second channel signals adjusted by the first and second direct filters comprise an amplitude and a time delay of each of the first and second channel signals.

20. The apparatus of claim 17, wherein the first binaural portion convolves the first channel signal with head related transfer functions measured between a first predetermined location and right and left ears of a listener to form a first right virtual signal and a first left virtual signal, the second binaural portion convolves the second channel signal with head related transfer functions measured between a second predetermined location symmetrical with the first predetermined location with respect to the listener and the right and left ears of the listener to form a second right virtual signal and a second left virtual signal, and the first and second right virtual signals are combined to form one of the first and second virtual signals and the first and second left virtual signals are combined to form the other one of the first and second virtual signals.

21. The apparatus of claim 17, wherein the crosstalk canceller comprises a crosstalk cancellation filter to cancel the crosstalk between the first and second virtual signals.

22. The apparatus of claim 21, wherein the crosstalk cancellation filter comprises an optimized IIR (infinite impulse response) filter.

23. The apparatus of claim 22, wherein the widening filter has coefficients determined by convolving coefficients of the head related transfer functions with coefficients of the crosstalk cancellation filter.

24. The apparatus of claim 23, wherein the widening filter and the first and second direct filters act as a panorama filter, and coefficients of the panorama filter are determined by adding the coefficients of the widening filter to coefficients of the first and second direct filters.

25. The apparatus of claim 17, further comprising: a phase inverter to invert a phase of one of the first and second channel signals when the first and second channel signals are the same.

26. The apparatus of claim 17, further comprising: a location ascertaining unit to ascertain a location of a listener; a table to store information corresponding to the location of the listener; and a controller to control the widening filter according to the information corresponding to the location of the listener.

27. The apparatus of claim 17, further comprising: a first speaker to generate sound corresponding to the filtered first channel signal and the first virtual signal; and a second speaker to generate sound corresponding to the filtered second channel signal and the second virtual signal.

28. A method of reproducing an input sound signal, the method comprising: filtering first and second channel signals of an input sound signal to adjust characteristics of the first and second channel signals; forming first and second virtual signals according to the input first and second sound signals and head related transfer functions; canceling crosstalk between the first and second virtual signals according to the head related transfer functions; and outputting the filtered first channel signal together with the first virtual signal, and the filtered second channel signal together with the second virtual signal.

29. The method of claim 28, wherein the filtering of the first and second channel signals comprises: adjusting the characteristics of the first and second channel signals according to the first and second virtual signals.

30. The method of claim 28, wherein the characteristics of the first and second channel signals comprise an amplitude and a time delay of each of the first and second channel signals.

31. The method of claim 28, wherein the forming of the first and second virtual signals comprise: convolving the first channel signal with head related transfer functions measured between a first predetermined location and left and right ears of a listener to form a first left virtual signal and a first right virtual signal; convolving the second channel signal with head related transfer functions measured between a second predetermined location symmetrical with the first predetermined location with respect to the listener and the left and right ears of the listener to form a second left virtual signal and a second right virtual signal; and combining the first and second left virtual signals to form one of the first and second virtual signals, and combining the first and second right virtual signals to form the other one of the first and second virtual signals.

32. The method of claim 28, wherein the canceling of the crosstalk between the first and second virtual signals comprises: passing the first and second virtual signals through a crosstalk cancellation filter having coefficients determined according to the head related transfer functions.

33. The method of claim 28, further comprising: inverting the phase of one of the first and second channel signals when the first and second channel signals are the same.

34. The method of claim 28, further comprising: ascertaining a location of a listener; and forming the first and second virtual signals and canceling cross-talk between the first and second virtual signals according to the location of the listener.

35. The method of claim 28, wherein the outputting of the filtered first channel signal together with the first virtual signal and the filtered second channel signal together with the second virtual signal comprises: generating sound corresponding to the first channel signal and the first virtual signal through a first speaker; and generating sound corresponding to the second channel signal and the second virtual signal through a second speaker.

36. An apparatus to reproduce an input sound signal, comprising: first and second direct filters to respectively filter first and second channel signals of an input sound signal according to first and second direct filter coefficients; a widening filter comprising: first and second binaural portions to form first and second virtual signals according to the input first and second channel signals and head related transfer functions (HTRF), and a crosstalk canceller to cancel crosstalk between the first and second virtual signals according to the head related transfer functions; and a first output terminal to output the filtered first channel signal convolved with the first virtual signal; and a second output terminal to output the filtered second signal convolved with the second virtual signal.

37. The apparatus of claim 36, wherein the first direct filter coefficient comprises a first display time and a first amplitude of the first channel signal, and the second direct filter coefficient comprises a second display time and a second amplitude of the second channel signal.