KR20010047701A - Method for implementing transaural filter for sound localization - Google Patents

Abstract

PURPOSE: A method for embodying a transaural filter for a space sound image lateralization is provided to minimize a distortion of a tone colour and a degradation of a tone quality CONSTITUTION: A method for embodying a transaural filter for a space sound image lateralization comprises an (a) step for establishing a ratio of a sound pressure transferred to the right ear and the left ear directly and a sound pressure transferred across from the bilateral symmetry type of the speaker, a (b) step for introducing the driving signal of the left speaker and the driving signal of the right signal applying a gain rate corresponding to the relative sound pressure ratio of the right and left ear and a delayed time corresponding to the different of reaching time between the right ear and the left ear and inverting 180°phase to the driving signal of speaker of the other side respectively, and a (c) step for applying the gain compensation rate to the obtained signal from the step (b).

Description

Method for implementing transaural filter for sound localization

The present invention relates to a method for implementing a transaural filter for spatial sound localization. More specifically, cross-talk is removed using the concepts of interaural level difference (ILD) and interaural time difference (ITD), which are spatial cognitive factors of human auditory organs, and the transoral filter is applied using masking characteristics of auditory organs. The present invention relates to a method for implementing a transoral filter capable of compensating amplitude and minimizing distortion and degradation of sound quality.

The sound felt when it reaches both ears directly from a single sound source located in space is defined by the sound pressure components {P _L , P _R } on the left and right. That is, sound pressure components {P _{L, i} , P _{R, i} ; i = 1, 2, ...}. However, when this spatial sound is implemented using two symmetrical speakers, the sound pressure component reaching both ears is a direct path sound pressure component {P, which directly transmits the sound generated from the left and right speakers to the left and right ears. _{LL, i} , P _{RR, i} ; i = 1, 2, ...} and cross-talk path sound pressure components in which the sound generated from the left and right speakers crosses each other to the right and left ears {P _{LR, i} , P _{RL, i} ; i = 1, 2, ...} respectively. Therefore, when the left and right sound transmission paths to the sound source under actual listening conditions drive the speaker, two types of direct sound pressure transmission paths and two types of left and right cross sound pressure transmission paths are involved. In particular, the left-right cross-negative pressure transmission components are called "cross-talk components", and the method of removing them is called "cross-talk removal technique." Cross-talk removal methods have been proposed in various ways, from the initial method proposed in the US patent in 1966 to the recently generalized methods. However, in the present invention, by using the classical free sound field model, two symmetrical speakers are used. We developed a method of converting a sound pressure transmission path of a species into two types of direct sound pressure transmission paths that are heard on both ears when listening to a single sound source.

An object of the present invention is to provide a method for implementing a transoral filter that can minimize the problems of sound quality degradation or tone distortion.

Another object of the present invention is to provide a method for implementing a transoral filter that can be easily implemented by recording technicians without specialized knowledge.

1 is a layout view of a listener and a sound source.

2 and 3 are graphs of trends in sound pressure ratio and time delay.

4 is a schematic diagram of the trans-oral filtering of the present invention.

5 is a graph relating to gain characteristics of the transoral filter of the present invention.

6 and 7 are graphs of equivalent gains in the n octave band.

In the present invention, the classical free sound field model was selected to minimize the tone distortion and the degradation of sound quality, which are problems of cross-talk removal using the conventional measurement head transfer function. Inverse filter due to numerous measurement errors (including loudspeakers and sound pressure sensor devices), resonance and notch components (acoustic characteristics of body parts such as head, ear wheels and shoulders) in the measurement head transfer function when using the free sound field model This is because problems can be avoided. In addition, the listener always recognizes the sound heard according to his acoustic sound pressure transmission characteristics according to his or her upper body, head, and outer ear as the most natural sound. In addition, the free sound field model focuses only on the reproduction of the sound field around both ears, not the sound pressure reproduction in the eardrum of both ears.

When the monopole source is driven with a signal of frequency f in the free sound field, the sound pressure at distance r is given by Equation (1).

In Equation 1, ρ ₀ = air density, jωq = volume acceleration of the sound source, k = wave count (ω / c ₀ ; ω = angular velocity, C ₀ = sound velocity). Characteristic value of sound source V = In this case, the sound pressure transfer function of the sound source with respect to the spatial position is shown in Equation 2.

Figure 1 shows the geometrical configuration of the speaker and the listener used in the present invention. R ₀ is the distance from the sound source to the center of the head, R ₁ and R ₂ are the direct path to the ear near the sound source and the cross-talk path distance from the far ear, and θ is the center of the head and the speaker The angle of and, and a is the radius of the listener's head. The sound pressure ratio g _c between the proximity path and the cross path and the sound pressure arrival time difference τ _c by the two path differences are defined as in Equation 3.

Where the sound pressure ratio g _c is always less than one.

2 and 3 show the change in the sound pressure ratio g _c and the sound pressure arrival time difference τ _c of both ears according to the installation conditions of the listener and the speaker. In a free sound field, the sound pressure is proportional to the inverse of the distance, so it is equal to the ratio of the simple distance. In FIG. 2, the sound pressure ratio decreases as the installation angle of the speaker increases, and the sound pressure ratio also decreases as the distance between the speaker and the listener increases. The sound pressure arrival time difference τ _c of the left and right ears increases as the installation angle of the speaker increases, but the influence of the distance is very small in FIG. 3. The information of the ratio of the sound pressure ratio and the time difference between the left and right ears of the listener to the speaker shown in FIGS. 2 and 3 is used in the implementation of the transaural filter.

When the driving sound source signals of the left and right speakers configured as shown in FIG. 1 are {S _L , S _R }, the direct path sound pressures of both ears from the sound pressure transfer function of Equation 2 are the same as Equation 4, and the cross talk path is a left and right cross component. Sound pressure is described by the linear model of Equation 5, respectively.

Therefore, the left and right ear sound pressures {P _L , P _R } are described by Equation 6 as the sum of sound pressure components of Equations 4 and 5.

Equation 6 is a linear model of the sound pressure around both ears when driving the left and right sound sources to two speakers. From the above acoustic model, when driving the left and right sound source signals to the speaker, when the sound pressure heard from both ears and the sound pressure of both ears directly from the sound source located in the space are the same, the spatial sound image localization using two speakers is performed. ) Can be implemented. As described above, it can be seen that spatial sound positioning using two symmetrical speakers involves the design of a driving sound source signal {S _L , S _R } according to the configuration of the speaker. Therefore, when the sound pressure on the right side of Equation 6 is the reference sound pressure heard by both ears when driving the spatial sound source, the speaker driving signal is expressed by Equation 7 below.

G (f) is a determinant value of the right side matrix of Equation 5, and is a complex function as follows.

Therefore, | G (f) | in Equation 7 is a gain value applied equally to the left and right speakers. Considering Equation 7, it is possible to understand a specific method of removing a cross-talk path in which the left and right speakers and both ears cross each other. For example, assuming that the sound pressure P _R = 0 of the right ear, the left speaker drives only a component proportional to P _L and the right speaker drives a signal proportional to (-g _c · e ^-jωτ ) × P _L. . At this time, the gain factor of the drive signal by right ^{_{(-g c · e -jωτ) ×}} P L is a non-relative negative pressure to the left and right of the ear speaker drive signal _{P L (Relative ILD) g c} (g c <1) Application and reach yanggwi It means the process of applying time delay τ _{c and} applying phase 180 degree transformation as much as time difference ITD. When the right and left signals obtained through these three processes (relative ILD, ITD, phase inversion) are applied to the speaker at the same time, only the driving signal component of the left speaker is heard in the left ear, and the cross transmission path of the left speaker in the right ear. The sound pressure caused by the sound pressure and the driving sound pressure of the right speaker overlap each other to remove the cross-transfer sound pressure component. Right drive signal in the same way also three kinds of processes described previously to remove the cross-sound pressure transmission elements are passed ears left (relative ILD, ITD, phase shift), that is, 180 degree phase shift, the gain factor g _c (g _c <1) and The respective time delays τ _c are applied to the left speaker. This calculation process is shown as shown in FIG.

The method of eliminating cross sound pressure transfer components involved in driving the symmetrical speaker mentioned above is only the first step of the transaural filter. The second step is gain equalization by the gain function | G (f) | which is equally applied to the left and right speaker drive signals as indicated in front of the speaker of FIG. (f) | is applied to the left and right speaker channels.

Since the phase information included in Equation 8 is equally applied to the left and right channels, it does not affect the recognition of the auditory organ. Therefore, in the present invention, | G (f) | is used instead of G (f) as in Equation (7). As shown in FIG. 1, the gain ratio | G (f) | of the left and right channels is determined according to the distance between the listener and the speaker and the angle {R ₀ , θ}.

The technique of "Stereo Dipole", developed by ISVR in the UK, chooses to realize G (f) directly as a digital gain processing function. In particular, the problem of implementing a phase term requires a lot of computation and complex processing. However, the stereo-dipole technique does not utilize the fact that the human auditory structure has the same interaural time difference (ITD) of zero when the phases of both ear pressures are in the same phase, and thus does not affect the spatial sound positioning.

Above we have described the concept of a transaural filter consisting of a first stage, the second stage that compensates for gain equal to the left and right, and the first stage of cross-pressure transfer component removal between two symmetrical speakers and both ears. This transaural filter uses symmetrical speakers to allow listeners to hear the sound field around both ears directly from a sound source located in a space of free sound field conditions. In other words, you get a transaural sound field reproduction, as if there is no speaker drive.

The following describes how to implement | G (f) |, a gain function of the left and right speakers. Fig. 5 shows the gain ratio | G (f) | for the angles θ = 5 ° and 30 °, respectively, of the sound source and the distance R ₀ = 2m, respectively.

The maximum value of gain factor is | G | _max = g _c / (1-g _c ² ) and the minimum value is | G | _min = g _c / (1 + g _c ² ). As shown in FIG. 5, when the distance R ₀ = 2m and the speaker angles θ = 5 ° and 30 °, g _c (5 °) = 0.9914 and g _c (30 °) = 0.9577. When the listener and speaker are at an angle of 5 ° | G | _max = 35.2 dB, | G | _min = -6 dB and when the angle is 30 ° | G | _max = 21.3 dB, | G | _min = -6 dB, respectively. It can be seen from FIG. 5 that the magnitude and shape of these peak values show a very sharp difference depending on the configuration of the listener and the speaker. First of all, we need a lot of gain compensation at low frequency, and the smaller the speaker angle, the more the size and bandwidth of the gain compensation. As the angle of the speaker increases, the size and bandwidth of the gain compensation decrease, but the number of peaks increases rapidly.

The most basic problem in the implementation of the transaural filter is the effective compensation of the gain function shown in FIG. 5, but the existing transaural filter techniques including the stereo dipole technique do not provide an effective solution to the above problem. As shown in FIG. 5, when the angle of the most widely used listener and speaker is 30 °, the frequency characteristic of the gain function | G (f) | with 11 peaks in the audible range up to 20 kHz is FIR or When implemented with an IIR digital filter, this is a major factor that can cause tonal distortion and degradation. In addition, there is a problem that a FIR or IIR digital filter that satisfies the frequency characteristic of the gain function | G (f) | having 11 peaks is maximized. In particular, when implementing these digital filters, it is necessary to select the optimal number of filter taps or the number of poles and zeros, and to prevent spatial sound distortion and degradation of sound quality according to the phase characteristics of the filter.

In the present invention, in order to solve a gain compensation technique that satisfies the frequency characteristic of the gain function | G (f) | having a large number of peaks, the equivalent bandwidth gain of the finite bandwidth is calculated using the "masking auditory characteristic". We adopt a method of performing gain compensation. Let low-lower frequency be f _L and high-frequency upper limit f _{H for} the center frequency f _c of the band of interest. At this time, the equivalent band gain value is defined as in Equation (9).

Selection of the center frequency, upper and lower limit bands of Equation 9 may select an auditory structure model such as octave, 1 / 3-octave, critical band, or recent Gammatone. 6 and 7 show the results for the 1 / 3-octave band, the most widely used acoustic technicians.

As can be seen from the results of FIG. 5, the largest characteristic of the gain function of FIGS. 6 and 7 using the equivalent bandwidth can be seen that first, a relatively large gain compensation is required in the low frequency band. The amount of compensation decreases as the angle of the speaker increases, and the bandwidth of the low frequency compensation band increases as the angle of the speaker decreases. It can be seen that the low frequency compensation value according to the distance does not show a big difference. However, it can be seen that the low frequency gain compensation value is relatively sensitive to the magnitude of the speaker installation angle. In particular, it can be seen that the compensation of the high frequency region is also increased when the speaker installation angle is 30 ° than when the speaker installation angle is 5 °.

In particular, in the case of 30 °, the gain ratio in the high frequency region shows an increase in gain compensation amount due to many maximum peaks of the gain function | G (f) | as shown in FIG. Many of the minimum peaks of the gain function | G (f) | in the high frequency region are buried by the maximum peak value, i.e. their effect is minimized because of the "masking" effect. In fact, many of the maximum and minimum peaks in these high frequency ranges involve many problems when implementing the gain function | G (f) | with an FIR or IIR filter. The digital filter implementation technique having such a problem is accompanied by deterioration of sound quality and distortion of timbre.

An equivalent band-specific gain value obtained by Equation 8 {EQ (f _c ); The desired gain ratio can be compensated for by setting the band-specific gain of the existing equalizer according to 20 Hz <= f _c <= 20 kHz}. When the left and right gain ratio compensation does not exactly match the left and right phase characteristics of the existing equalizer it is inevitable to influence the phase of the speaker input signal. Since the phase influence of the equalizer is converted into the finite time delay amount of the left and right sides in the spatial sound positioning, the spatial sound positioning can be hindered. In the present invention, a zero-phase compensation technique, that is, a method of applying half of the gain in the forward direction with respect to the time axis and then inverting the signal back to the inverse of the time axis, applies the gain of the other half. The method may enable zero-phase gain compensation to completely eliminate the effects of phase with any digital equalizer.

The implementation of the transoral filter described above can be easily implemented using conventional hardware such as volume, time delay, phase inverter, and digital or analog equalizer, and the hardware implementation procedure can be implemented using sound editing software. have.

The invention consists of the direct sound and the cross sound pressure ratio g cross-talk component removal process of the left and right speakers Using a gain volume and the time delay to set the _c and and the zero-phase gain compensation stage using the existing digital equalizer of. This implementation provides the advantages that domestic sound design and recording technicians can easily implement because they do not require the expertise of digital signal processing technology or related DSP hardware. In particular, the zero-phase gain compensator using the equalizer has problems of sound quality degradation and timbre distortion caused by the digital filtering technique required to implement the gain function G (f) of the conventional transaural filter and the approximation and phase distortions associated with the related DSP. It has the advantage of minimizing.

Claims (4)

(a) setting the ratio of sound pressure delivered cross-directly and sound pressure delivered directly to the left and right ears in a symmetrical speaker; (b) Applying the gain ratio as much as the relative sound pressure ratio of the left and right ears and the time delay as much as the arrival time difference between both ears, to the left speaker driving signal and the right speaker driving signal, and inverting the 180 ° phase to the opposite speaker driving signal, respectively. Authorized; And, (c) applying the same gain compensation rate to the signal obtained in step (b) again; Transoral filter implementation method for spatial sound phase, characterized in that consisting of. The method of claim 1, wherein the gain compensation ratio is in the range of the lower limit frequency f _L and the higher frequency limit f _U with respect to the center frequency f _c . 12. A method for implementing a transoral filter for spatial sound localization, comprising using an equivalent band-specific gain defined by a value. [4] The transoral for spatial acoustic stereotype of claim 2, wherein the gain value of each equivalent band is applied half of the gain rate forwardly with respect to the time axis, and then reverses the signal to the inverse of the time axis and applies the gain rate of the other half. How to implement a filter. The method of any one of claims 1 to 3, wherein the implementation of the transoral filter is implemented using hardware such as volume, time delay, phase inverter and digital or analog equalizer, or using sound editing software. A method for implementing a transoral filter for spatial sound phase, characterized in that.