Room Impulse Response Generator

dr.ir. Emanul A.P. Habets e

[email protected]

September 20, 2010

Abstract This report provides a short overview of dierent methods that can be used for simulating room acoustics and focusses on the imagemethod that was proposed by Allen and Berkley in 1979. The image method is probably one of the methods most commonly used in the acoustic signal processing community, and will therefore be discussed in more detail. A mex-function, which can be used in MATLAB, has been created to generate multichannel Room Impulse Responses using the image method. This function enables the user to control the reection order, room dimension and microphone directivity.

Introduction

Many people who are working in the eld of acoustic signal processing reach a point where they want to simulate room acoustics. This report gives a short overview of dierent methods that can be used for simulating room acoustics. The image method [1], which was proposed by Allen and Berkley in 1979, is probably one of the methods most commonly used in the acoustic signal processing community. This method is closely related to the so-called wave equation and its frequency domain counterpart called the Helmholtz equation. These equations describe the propagation of acoustic waves through uids (gas or liquid). A mex-function, which can be used in MATLAB, has been created to generate multichannel Room Impulse Responses (RIR) using the image method. This function enables the user to control the reection order, room dimension and microphone directivity. Another advantage of this mex-function, compared to a standard MATLAB function, concerns the computation time. We have found that our mex-function was around 100 times faster than the standard MATLAB code on our test machine. This report is organized as follows. The wave equation and the Helmholtz equation are discussed in Sections 2 and 3 respectively. In Section 4 a short overview of dierent methods that can be used to simulate room acoustics is provided. Allen and Berkleys method is described in Section 5. A short description of the mex-function implementation is provided in Section 6. Some examples are presented in Section 7. Finally, a short summary is provided in Section 8. Please note and respect the copyrights in Section 11.

Wave Equation

In principle, any complex sound eld can be considered as a superposition of numerous simple sound waves (e.g., plane waves), and their propagation within a room can be con Copyright 2003-2010 by E.A.P. Habets. The text of the report may be reproduced free of charge in any format or media without requiring specic permission. The source of the material must be acknowledged and the title of the document must be included when being reproduced as part of another publication or service.

sidered linear if the properties of the medium in which the waves travel are assumed to be homogeneous, at rest, and independent of wave amplitude [2]. In physics, the wave equation governs the propagation of waves through uids (gas or liquid). The form of the equation is a second order partial dierential equation. The equation describes the evolution of velocity potential or sound pressure p(r, t) as a function of position r = [x, y, z] and time t. For a homogeneous medium undergoing inviscid uid ow, one can linearize the equations governing the dynamic behaviour of the uid, namely the Eulers equation (i.e., Newtons 2nd law applied to uids), the continuity equation, and the state equation, to obtain the wave equation 1 2 p(r, t) 2 = 0, (1) p(r, t) 2 c t2 where
2

2 2 2 + 2+ 2 x2 y z

(2)

is the Laplacian expressed in the Cartesian coordinates (x, y, z), and c is the speed of sound. The wave equation accurately describes the pressure in the sound eld provided |p(r, t)| 0 c2 , where 0 is the density of the propagation medium at equilibrium. In practice, two types of inhomogeneities occur in the medium: i) scalar inhomogeneities (spatial distribution of sound speed and air density), e.g., due to temperature variations in the medium, and ii) vector inhomogeneities (spatial distribution of particle mean velocity), e.g., due to the presence of fans or an air conditioning. However, the eects of these inhomogeneities are so small that they can often be ignored in room acoustics. In order to calculate the sound eld emanating from a source in a specic room we need an additional source function in (1) and boundary conditions that describe the sound reection and absorption at the walls. Let s(r, t) denote the source function, then the wave equation is given by 1 2 p(r, t) 2 p(r, t) 2 = s(r, t). (3) c t2

Helmholtz Equation

Let us consider the wave equation in the frequency domain. The Fourier transform is dened as P (r; ) F{p(r, t)}() = p(r, t) exp(it) dt, (4)

where exp(x) = ex and i = 1. By applying the Fourier transform to (1) the timeindependent Helmholtz equation is obtained, i.e.,
2

P (r; ) + k 2 P (r; ) = 0,

(5)

where k denotes the wave number that is related to the angular frequency and the wave length through 2 . k= = c If there is a harmonic disturbance which is producing the waves, for which the source function is given by s(r, t) = S(r; )eit , then the Helmholtz equation is given by
2

P (r; ) + k 2 P (r; ) = S(r; ).

(6)

For a unit-amplitude harmonic point source at position rs = [xs , ys , zs ] we have S(r; ) = (r rs ) = (x xs )(y ys )(z zs ), where () denotes the Kronecker delta function. The

partial dierential equation in (6) can be solved by rst solving the following inhomogeneous equation: 2 H(r, rs ; ) + k 2 H(r, rs ; ) = (r rs ), (7) where H(r, rs ; ) is the Room Transfer Function (RTF), or Greens function. For an arbitrary source function S(rs ; ) the desired source pressure can then be calculated using the following relation P (r; ) =
Vs

H(r, rs ; ) S(rs ; ) drs ,

(8)

where Vs denotes the source volume, and drs = dxs dys dzs is the dierential volume element at position rs . The sound pressure p(r, t) can now be obtained using the inverse Fourier transform of (8). The conventional way to solve (7) is to nd an orthogonal set of eigenfunctions associated with the Laplacian operator and then to expand H(r, rs ; ) as a sum of eigenfunctions. Specically, a function m (r; ) that satises the homogenous equation ( 2 +k 2 )m (r; ) = 0 over a certain interval and satises certain boundary conditions at the end of the interval, is called an eigenfunction. Subsequently, a general expression for the Greens function H(r, rs ; ) in an arbitrary sound eld can be obtained using the eigenfunctions: H(r, rs ; ) =
m=0

Cm (rs ; ) m (r; ),

(9)

where each coecient Cm (rs ; ) depends on the position of the sound source. The eigenfunctions depend on the boundary conditions imposed by the enclosed space.

Free Space Greens Function

For an omnidirectional point source in an unbounded space, i.e., free space, the Greens function is [2] exp(i r rs ) c , (10) H(r, rs ; ) = 4 r rs where denotes the Euclidean norm.

Classical Rectangular Room

In a rectangular room with physical dimensions (Lx , Ly , Lz ) and rigid perfectly reecting walls the eigenfunctions in Cartesian coordinates are [2] m (r) = cos mx x cos Lx my y cos Ly mz z , Lz (11)

where m = (mx , my , mz ) and mv {x, y, z} are nonnegative integers. Let us dene kv = mv /Lv for v {x, y, z}, then we can write (11) as m (r) = cos(kx x) cos(ky y) cos(kz z). (12)

The eigenfunctions are often referred to as modes and have a simple physical interpretation 2 2 2 2 as three-dimensional standing waves. The corresponding eigenvalues are km = kx + ky + kz . The solution for the inhomogeneous equation (7) for a classical rectangular room is [2] H(r, rs ; ) =
mM

m (r) (rs ) m , 2 m (k 2 km )

(13)

Computational Modelling of Room Acoustics

Wave based Modelling

Ray based Modelling

Statistical Modelling

Dierence Methods

Elements Methods

Ray Tracing

Image Source Method

SEA

FEM

BEM

Figure 1: Room acoustic models are based on sound rays (ray-based), on solving the wave equation (wave-based) or some statistical method [3]. where M = {(mx , my , mz ) : mx , my , mz N0 } denotes a set that contains all desired triples m and m is a normalization constant for the associated eigenvector dened by m (r) (r) dr = n m , for m = n; 0, otherwise, (14)

where V = {(x, y, z) : 0 x Lx , 0 y Ly , 0 z Lz } is the entire space of the room and dr = dx dy dz is the dierential volume element at position r. Equation (13) reveals the frequency domain structure of the RTF. The eigenfrequencies m m , related to the eigenvalues through km = c , are also known as the resonance frequencies of the room. At each eigenfrequency m , the standing wave pattern of mode m resonates strongly. From (13) it can be seen that H(r, rs ; ) increases without bound as m . Room mode m (r) is said to be excited at eigenfrequency m (i.e., m (r) makes a large contribution to sound pressure at this frequency). All rooms possess distinct resonances at low frequencies. However, in practical rooms, where walls are non-rigid and nitely absorbing, eigenvalues km have imaginary components that provide damping of resonance m m modes [2]. In that case km = c + i c , where m denotes the damping constant (Q-factor). Assuming that m m , (13) results in H(r, rs ; ) = c2
mM

m (r) (rs ) m . 2 m ( 2 m 2im m )

(15)

The inverse Fourier transform of the frequency response of the room described by (13) leads to a RIR, h(r, rs , t). The form of (13) justies the use of some well-known modelling techniques used in signal processing such as, for example, the pole-zero model.

Simulating Room Acoustics

Mathematically the sound propagation is described by the wave equation. An impulse response from a source to a microphone can be obtained by solving the wave equation. Since it can seldom be expressed in an analytic form the solution must be approximated. There are three main modelling methods, as illustrated in Figure 1, viz., wave-based, ray-based and statistical [3]. The ray-based methods, such as the ray-tracing [4] and the imagesource method [1], are the most often used. The wave-based methods, such as the Finite

Element Method (FEM), Boundary Element Method (BEM) [5, 6] and Finite-Dierence Time-Domain (FTDT) [7] methods, are computational more demanding. In real-time auralization1 the limited computation capacity requires simplications. A frequently used simplication consists of modelling the direct path and early reections individually and the late reections by recursive digital lter structures. The statistical modelling methods, such as the Statistical Energy Analysis, have been widely used in aerospace, ship and automotive industry for high frequency noise analysis and acoustic designs. They are not suitable for auralization purposes since those methods do not model the temporal behaviour of a sound eld. Wave-based methods The most accurate results can be achieved by wave-based methods. An analytical solution for the wave equation can be found only in extremely simple cases such as a rectangular room with rigid walls. Therefore, numerical methods such as FEM and BEM [5,6] are often used. The main dierence between these two element methods is in the element structure. In FEM, the space is divided into volume elements, while in BEM only the boundaries of the space are divided into surface elements. The elements interact with each other according to the basics of wave propagation. The sizes of these elements have to be much smaller than the size of the wavelength for every particular frequency. At high frequencies, the required number of elements becomes very high, resulting in a large computational complexity. Therefore, these methods are suitable only for low frequencies and small enclosures. Another method for room acoustics simulation is provided by the FDTD method [7, 8]. The main principle of this method is that derivatives in the wave equation are replaced by corresponding nite dierences. The FDTD method produces impulse responses that are better suited for auralization than FEM and BEM. The main benet of the element methods over FDTD methods is that one can create a denser mesh structure where required, such as locations near corners or other acoustically challenging places. In all wave-based methods, the most dicult part is the denition of the boundary conditions and geometrical description of the objects. Typically a complex impedance is required, but it is hard to nd that data in existing literature. Ray-based methods The ray-based methods are based on geometrical room acoustics [2]. The most commonly used ray-based methods are the ray-tracing [4] and the image method [1]. The main dierence between these methods is the way the reection paths are calculated [3]. To model an ideal impulse response from a source to a receiver all possible sound reection paths, commonly called rays, should be discovered. In ray-tracing methods the sound power emitted by a sound source is described by a nite number of rays. These rays propagate through space and are reected after every collision with the room boundaries. During that time, their energy decreases as a consequence of the sound absorption of the air and of the walls involved in the propagation path. When the rays reach the receiver, an energy calculation process is performed. When all rays are processed the impulse response is obtained. Rays can be selected from a set of randomly distributed angles, uniformly distributed angles or from a restricted set of angles. Due to this the ray-tracing methods are by no means exhaustive, whereas the image method nds all the rays. However, while the image method is limited to geometries that are formed by planer surfaces the ray-tracing method can be applied to geometries that are formed by arbitrary surfaces.
1 Auralization is the process of rendering audible, by physical or mathematical modelling, the sound eld of a source in a space, in such a way as to simulate the binaural listening experience at a given position in the modelled space.

S 1 0 1 0 1 0 R

S 11 00 11 00 11 00

Figure 2: Path involving one reection obtained using one image.

It should be mentioned that all ray-based methods are based on energy propagations. This means that all eects involving phase dierences such as refraction or interference are neglected. This is admissible if the sound signals of interest are not sinusoids or other signals with small frequency bandwidth but are composed of many spectral components covering a wide frequency range. Then it can be assumed that constructive and destructive phase eects cancel each other when two or more sound eld components superimpose at a point, and the total energy in the considered point is simply obtained by adding their energies. Components with this property are often referred to as mutually incoherent [9].

Allen and Berkleys Image Method

The image model can be used to simulate the reverberation in a room for a given source and microphone location, and is discussed in Section 5.1. Using the image method Allen and Berkley [1] developed an ecient method to compute a Finite Impulse Response (FIR) that models the acoustic channel between a source and a receiver in rectangular rooms. The image method and some additional renements will we discussed in Section 5.2. The close relation of the image method and the Helmholtz equation is shown in Section 5.3.

5.1

Image Model

Figure 2 shows a sound source S located near a rigid reecting wall. At destination D two signals arrive, one from the direct path and a second one from the reection. The path length of the direct path can be directly calculated from the known locations of the source and the destination. Also shown is an image of the source, S , located behind the wall at a distance equal to the distance of the source from the wall. Because of symmetry, the triangle SRS is isosceles and therefore the path length SR + RD is the same as S D. Hence, to compute the path length of the reected path, we can construct an image of the source and compute the distance between destination and image. Also, the fact that we are computing the distance using one image means that there was one reection in the path. Figure 3 shows a path involving two reections. The length of this path can be obtained from the length of S D. In Figure 4 the length of a path involving three reections is obtained from the length of S D. These gures can also be extended to three dimensions to take into account reections from the ceiling and the oor. In general the path lengths (and thus the delays) of reections can be obtained by computing the distance between the source images and the destination. The strength of the reection can be obtained from the path length and the number of reections involved in the path. The number of reections involved in the path is equal to the level of images that was used to compute the path.

S 1 0 1 0

S 11 00 11 00

1 0 1 0 S

Figure 3: Path involving two reections obtained using two images.

1 0 1 0 1 0

1 0 1 0 1 0

1 0 1 0

S 11 00 11 00

1 0 1 0 1 0

Figure 4: Path involving three reections obtained using three images.

5.2

Image Method

Consider a rectangular room with length, width and height given by Lx , Ly and Lz . Let the sound source be at a location represented by the vector rs = [xs , ys , zs ] and let the microphone be at a location represented by the vector r = [x, y, z]. Both vectors are with respect to the origin, which is located at one of the corners of the room. The relative positions of the images measured with respect to the receiver position and obtained using the walls at x = 0, y = 0 and z = 0 can be written as Rp = [(1 2q)xs x, (1 2j)ys y, (1 2k)zs z] . (16)

Each of the elements in the triple p = (q, j, k) can take on values 0 or 1, resulting in eight dierent combinations that specify a set P, i.e., P = {(q, j, k) : q, j, k {0, 1}}. When the value of p is 1 in any dimension, then an image of the source in that direction is considered. It should be noted that some of these images correspond to higher order reections. To consider all images, we add the vector Rm to Rp where Rm = [2mx Lx , 2my Ly , 2mz Lz ], (17)

where mx , my , and mz are integer values. Each of the elements of the triple m = (mx , my , mz ) takes on values from N to +N . The reection order related to an image at the position Rp + Rm + r is given by Op,m = |2 mx q| + |2 my j| + |2 mz k|. The distance between any source image and the microphone can be written as d = Rp + Rm . (19) (18)

The time delay of arrival of the reected sound ray corresponding to any source image can be expressed as Rp + Rm d , (20) = = c c where c denotes the sound velocity in meters per second. The impulse response for this source and microphone location can now be written as h(r, rs , t) =
pP mM |m |m |m |m |m |m x1 x q| x2 x | y1 y j| y2 y | z1 z k| z2 z |

(t ) , 4d

(21)

where M = {(mx , my , mz ) : N mx , my , mz N } denotes a set that contains all desired triples m. The quantities x1 , x2 , y1 , y2 , z1 and z2 are the reection coecients of the six walls. Note that the walls at v = 0 with v {x, y, z} correspond to v1 , and that the walls at v = Lv with v {x, y, z} correspond to v2 . The elements of the triple p are 0 or 1, which means that there are 8 dierent combinations, (0, 0, 0) to (1, 1, 1). The elements of the triple m range from N to +N , which means that there are (2N + 1)3 combinations. Therefore, for a given N , this method computes 8(2N + 1)3 dierent paths. The delays of the impulses corresponding to these paths are computed using (20) and the strengths of these impulses are multiplied by reection coecients as many times as there are reections. Once the impulse response has been computed this way, the source signal can be convolved with the impulse response to simulate the signal picked up by the microphone. An important consideration while simulating the discrete version of this impulse response using a computer is that the delays given by (20) do not always fall at sampling instants. Ideally, the discrete version of (21) is given by h(r, rs , n) =
pP mM |m |m |m |m |m |m x1 x q| x2 x | y1 y j| y2 y | z1 z k| z2 z |

LPF{(n fs )} , 4d

(22)

1.2 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 0 2 4 6 8 10

n [samples]

Figure 5: Comparison of the shifted and low-pass impulse method.

where fs is the sampling frequency and LPF{} denotes a theoretically perfect Low-Pass Filter with cut-o frequency fs /2. In [1] the time-of-arrival (in samples) was shifted to the nearest integer value. Hence, the following approximation was made LPF{(n fs )} (n round{ fs }). (23)

Although this distortion can be ignored in many applications, for multiple microphone systems that are sensitive to inter-microphone phase, correct simulation of arrival time relationships is critical. One way to reduce this problem is to compute the discrete impulse response at a much higher sampling frequency, decimate the impulse response to the original sampling frequency, and convolve the source signal with it. Peterson suggested another modication to the image method [10]. In this approach, each impulse in (21) is replaced by the impulse response of a Hanning-windowed ideal low-pass lter of the form LPF (t) =
1 2

1 + cos

2t Tw

sinc (2fc t)

w for T2 < t < otherwise

Tw 2

(24)

where Tw is the width (in time) of the impulse response and fc is the cut-o frequency of the low-pass lter. For the simulations performed in this report Tw was set to 4 ms and fc was set to the Nyquist frequency. Each impulse (t ) in (21) is rst replaced by LPF (t ) and subsequently sampled. By doing this, true delays of arrival of the reected signals are simulated accurately even at the original low sampling frequency. A comparison of both methods is depicted in Figure 5, where the delay was set to 4.8 samples. Squares indicate sample values produced by Allen and Berkleys shifted impulse method and circles indicate values produced by Petersons low-pass impulse method. The solid line shows the central portion of the continuous-time low-pass impulse function. The other consideration while simulating reverberation for a room is the duration of reverberation or the reverberation time. Formally, the reverberation time is dened as the time required for the intensities of reected sound rays to be down 60 dB from the direct path sound ray. An empirical formula, known as Sabin-Franklins formula [11] can be used to relate the reverberation time RT60 by, RT60 = 24 ln(10) V c
6 i=1 2 Si (1 i )

(25)

where V denotes the volume of the room, and i and Si denote the reection coecient and the surface of the ith wall, respectively.

10

5.3

Relation with the Helmholtz Equation

The image model described in Section 5.1 yields an intuitive explanation of the imagemethod. In this section we will proof the relation between the solution of the Helmholtz equation (7) for a classical rectangular room and (21). The solution of the Helmholtz equation for a rectangular room was given in (13). Let us take all images into account by dening < mv < for v {x, y, z}. By expanding the cosines of the eigenfunctions using exponentials, and by using the fact that m = V m, we can write (13) as H(r, rs ; ) = where km = 1 8V
pP mM

exp(ikm Rp ) , k 2 km 2

(26)

mx my mz , , = [kx , ky , kz ], Lx Ly Lz

(27)

and Rp represents the eight vectors given by (16). Now we proceed along the same lines as Allen and Berkley in [1]. Using the property of the delta function on kx , ky , and kz , i.e.

(k a)f (k) = f (a),

(28)

we may rewrite (26) in integral form H(r, rs ; ) = 1 8V exp(ik Rp ) k2 k 2

mv =

pP

mM

(k km ) dk ,

(29)

where k = [kx , ky , kz ]. By Fourier series analysis one may show

mv =

kv

mv Lv

Lv

exp(i2Lv mv kv )

for v {x, y, z}.

(30)

Therefore, we can rewrite (29) as H(r, rs ; ) = 1 (2)3

pP mM

exp(ik (Rp + Rm )) dk , k2 k 2

(31)

where Rm is the vector dene in (17). The integral in (31) is just a plane wave expansion for a point source in free space since exp(ik R ) 1 = 4 R (2)3 Finally, using (31) and (32) we obtain H(r, rs ; ) =
pP mM

exp(ik R) dk . k2 k 2

(32)

exp(ik Rp + Rm ) . 4 Rp + Rm

(33)

Taking the inverse Fourier transform2 of (33) the RIR is obtained h(r, rs , t) =
pP mM

(t Rp + Rm /c) 4 Rp + Rm (t ) . 4d

(34) (35)

=
pP mM
2 Note

that the inverse Fourier transform is dened as p(t) =

P () exp(it) d.

11

By taking into account the reection coecients of the six walls we obtain h(r, rs , t) =
pP mM |m |m |m |m |m |m x1 x q| x2 x | y1 y j| y2 y | z1 z k| z2 z |

(t ) , 4d

(36)

which is equivalent to (21).

Implementation

The image method as discussed in the previous section has been implemented as a MATLAB mex-function and was written in C++. The resulting Dynamic-Link-Library (DLL) can easily be used within MATLAB as a standard MATLAB function. The C++ implementation is much faster than the equivalent MATLAB implementation. The source-code can be found in Appendix A. The function rir generator is dened as follows: function [h, beta_hat] = rir_generator(c, fs, r, s, L, beta, nsample, mtype, order, dim, orientation, hp_filter); Input parameters: Description Parameter c sound velocity in m/s. fs sampling frequency in Hz. r M x 3 matrix specifying the (x,y,z) coordinates of the receiver(s) in m. s 1 x 3 vector specifying the (x,y,z) coordinates of the source in m. L 1 x 3 vector specifying the room dimensions (x,y,z) in m. 1 x 6 vector specifying the reection coecients [x1 x2 y1 y2 z1 z2 ] beta or beta = Reverberation Time (RT60 ) in seconds. Optional input Parameter nsample mtype order dim orientation hp lter parameters: Description number of samples to calculate. type of microphone that is used [omnidirectional, subcardioid, cardioid, hypercardioid, bidirectional]. maximum reection order. room dimension (2 or 3). direction in which the microphone is pointed, specied using azimuth and elevation angles in radians. use false to disable high-pass lter. Default value RT60 fs omnidirectional -1 3 [0 0] true

Output parameters: Parameter Description h M x nsample matrix containing the calculated room impulse response(s). beta hat In case a reverberation time is specied as an input parameter the corresponding reection coecient is returned. Multi-Channel Support In case more than one receiver position is specied the function rir generator will calculate all RIRs at once. Reverberation Time versus Reection Coecients The reection coecients in (21) can be specied using the parameter beta. In case beta consists of one element the

12

program assumes a reverberation time (in seconds) is specied. The corresponding average reection coecient is calculated using (25) and will be returned using the output parameter beta hat. Reection Order and Room Dimension In order to control the complexity of the generated RIR one can control the maximum reection order using the parameter order . In case the order is chosen -1 (default value) the maximum amount of reections, given the desired length of the RIR, is calculated. The dimension of the room can be set using the parameter dim. This value can either be 2 or 3 (default value). Microphone Directivity The microphones directionality, or polar pattern, can also be taken into account. Dierent kinds of polar patterns are implemented and can be chosen using the parameter mtype. The signal attenuation A(), where denotes the direction of arrival, is calculated using the following standard formula: A() = + (1 ) cos () . (37)

The polar pattern is controlled by , see Table 1. The resulting polar patterns for the Omnidirectional, Cardioid, Hypercardioid and Bidirectional microphone are depicted in Figure 6. Directivity Pattern Omnidirectional (Monopole) Subcardioid Cardioid Hypercardioid Bidirectional (Dipole) 1 0.75 0.5 0.25 0

Table 1: Supported polar patterns and corresponding values for . The angle in which the microphone is pointing can be adjusted with the parameter orientation. By default the microphone points towards the positive x-axis. The microphones directionality takes the azimuth and elevation of the received reections into account.

13

120 150 1 180 150

Omni-directional 90 60 30 150 1 0 30 180 150

120

Cardioid 90

60 30

0.5

0.5

0 30

120

120 150 1 180 150

Hyper-cardioid 90 60 30 150 1 0 30 180 150

90

60

120

120

Bi-directional 90

90

60

60 30

0.5

0.5

0 30

120

90

60

120

90

60

Figure 6: Polar plots of four dierent microphone polar patterns.

Examples

In this section some basic and more complex examples are presented in the form of a MATLAB script.
c = 340; f s = 16000; r = [2 1.5 2 ] ; s = [2 3.5 2 ] ; L = [5 4 6 ] ; beta = 0 . 4 ; n = 4096; h = rir generator (c , fs , r, % % % % % % % s, S o u n d v e l o c i t y (m/ s ) Sample f r e q u e n c y ( samples / s ) R e c e i v e r p o s i t i o n [ x y z ] (m) S o u r c e p o s i t i o n [ x y z ] (m) Room d i m e n s i o n s [ x y z ] (m) Reverberation time ( s ) Number o f s a m p l e s L, beta , n);

Example 1: Simple example to generate one RIR.

0.04 0.03 0.02 0.01 0 -0.01 500 1000 1500 2000 2500 3000 3500 4000

n [samples]

Figure 7: Output of Example A.1.

14

c = 340; f s = 16000; r = [2 1.5 2 ] ; s = [2 3.5 2 ] ; L = [5 4 6 ] ; beta = 0 . 4 ; n = 1024; mtype = o m n i d i r e c t i o n a l ; order = 2; dim = 3 ; o r i e n t a t i o n =[0 0 ] ; h p f i l t e r =1; h = rir generator (c , fs , r,

% % % % % % % % % % % % s,

S o u n d v e l o c i t y (m/ s ) Sample f r e q u e n c y ( samples / s ) R e c e i v e r p o s i t i o n [ x y z ] (m) S o u r c e p o s i t i o n [ x y z ] (m) Room d i m e n s i o n s [ x y z ] (m) Reverberation time ( s ) Number o f s a m p l e s Type o f m i c r o p h o n e Reflection order Room d i m e n s i o n Microphone o r i e n t a t i o n ( rad ) E n a b l e h i g h p a s s f i l t e r L, beta , n, mtype , o r d e r , dim , o r i e n t a t i o n , h p f i l t e r ) ;

Example 2: Generate one RIR.

0.04 0.03 0.02 0.01 0 -0.01 100 200 300 400 500 600 700 800 900 1000

n [samples]

Figure 8: Output of Example A.2.

c = 340; f s = 16000; r = [2 1.5 2 ; 1 1.5 2 ] ; s = [2 3.5 2 ] ; L = [5 4 6 ] ; beta = 0 . 4 ; n = 4096; mtype = o m n i d i r e c t i o n a l ; o r d e r = 1; dim = 3 ; o r i e n t a t i o n =[0 0 ] ; h p f i l t e r = 1; h = rir generator (c , fs , r,

% % % % % % % % % % % % s,

S o u n d v e l o c i t y (m/ s ) Sample f r e q u e n c y ( samples / s ) Receiver positions [ x 1 y 1 z 1 ; x 2 S o u r c e p o s i t i o n [ x y z ] (m) Room d i m e n s i o n s [ x y z ] (m) Reverberation time ( s ) Number o f s a m p l e s Type o f m i c r o p h o n e 1 e q u a l s maximum r e f l e c t i o n o r d e r ! Room d i m e n s i o n Microphone o r i e n t a t i o n ( rad ) E n a b l e h i g h p a s s f i l t e r L, beta , n,

y 2

z 2 ]

(m)

mtype , o r d e r , dim , o r i e n t a t i o n , h p f i l t e r ) ;

Example 3: Generate multiple RIRs.

c = 340; f s = 16000; r = [2 1.5 2 ] ; s = [2 3.5 2 ] ; L = [5 4 6 ] ; n = 4096; beta = 0 . 4 ; mtype = h y p e r c a r d i o i d ; o r d e r = 1; dim = 3 ; o r i e n t a t i o n =[ p i /2 0 ] ; h p f i l t e r = 1; h = rir generator (c , fs , r,

% % % % % % % % % % % % s,

S o u n d v e l o c i t y (m/ s ) Sample f r e q u e n c y ( samples / s ) R e c e i v e r p o s i t i o n [ x y z ] (m) S o u r c e p o s i t i o n [ x y z ] (m) Room d i m e n s i o n s [ x y z ] (m) Number o f s a m p l e s Reverberation time ( s ) Type o f m i c r o p h o n e 1 e q u a l s maximum r e f l e c t i o n o r d e r ! Room d i m e n s i o n Microphone o r i e n t a t i o n ( rad ) E n a b l e h i g h p a s s f i l t e r L, beta , n, mtype , o r d e r , dim , o r i e n t a t i o n , h p f i l t e r ) ;

Example 4: Generate one RIR using a hypercardioid microphone.

15

0.04 0.03 0.02 0.01 0 -0.01 500 1000 1500 2000 2500 3000 3500 4000

n [samples]

Figure 9: Output of Example A.4. A reverberant signal can now be created by ltering the anechoic signal with the generated RIR as shown in Example 5.
reverberant signal = f f t f i l t (h , c l e a n s i g n a l ) ;

Example 5: Generate reverberant signal.

Summary

In this report we presented a short overview of dierent methods for simulating room acoustics. We discussed the well-known image method, proposed by Allen and Berkley [1], in more detail. An ecient implementation of the image method was developed in the form of a MATLAB mex-function written in C++. Some example scripts were presented to demonstrate the use of this function. Our implementation incorporates some novel features which allows the user to control the complexity of the RIR and the directivity pattern of the receiver.

Acknowledgements

The author expresses his thanks to J. van de Laar and N.D. Gaubitch for proofreading the initial version of this manuscript and testing the mex-function, and to D. Jarrett for fruitful discussions and comments that gave rise to the latest version of the RIR Generator and this tutorial.

10

Contact Information

If you have any comments and/or suggestions you can contact the author at the following address: Mail: d.ir. Emanul A.P. Habets e Mailstop: EEE-CSP Department of Electrical and Electronic Engineering Imperial College London Exhibition Road London SW7 2AZ United Kingdom [email protected]

E-mail:

16

11

Copyrights

Copyright 2003-2010 E.A.P. Habets, The Netherlands. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA

A
/ Program

Source Code
: : Room Impulse Response Generator t o one o r more image method [ 1 , 2 ] .

Description

C o m p u t e s t h e r e s p o n s e o f an a c o u s t i c s o u r c e m i c r o p h o n e s i n a r e v e r b e r a n t room u s i n g t h e

[ 1 ] J . B . A l l e n and D . A . B e r k l e y , I m a g e m e t h o d f o r e f f i c i e n t l y s i m u l a t i n g s m a l l r o o m a c o u s t i c s , J o u r n a l A c o u s t i c S o c i e t y o f America , 6 5 ( 4 ) , A p r i l 1979 , p 9 4 3 . [ 2 ] P .M. P e t e r s o n , Simulating the response of multiple microphones to a s i n g l e a c o u s t i c s o u r c e i n a r e v e r b e r a n t room , J o u r n a l A c o u s t i c S o c i e t y o f America , 8 0 ( 5 ) , November 1 9 8 6 . Author Version History : : : dr . i r . E.A.P. Habets ( ehabets@dereverberation . org )

2.0.20100920 1.0.20030606 1.1.20040803 1.2.20040312 1.3.20050930 1.4.20051114 1.5.20051116 1.6.20060327 1.7.20060531 1.8.20080713 1.9.20090822 2.0.20100920 Initial version + Microphone d i r e c t i v i t y + Improved phase accuracy [ 2 ] + Reflection order + R e v e r b e r a t i o n Time + S u p p o r t s m u l t i c h a n n e l s + H i g h p a s s f i l t e r [ 1 ] + Microphone d i r e c t i v i t y c o n t r o l + Minor i m p r o v e m e n t s + Minor i m p r o v e m e n t s + Minor i m p r o v e m e n t s + 3D m i c r o p h o n e d i r e c t i v i t y c o n t r o l + C a l c u l a t i o n o f t h e s o u r c e i m a g e p o s i t i o n c h a n g e d i n t h e c o d e and t u t o r i a l . This ensures a proper response to r e f l e c t i o n s in case a d i r e c t i o n a l microphone i s used . Habets , The Netherlands .

Copyright

(C)

2003 2010 E . A . P .

T h i s program i s f r e e s o f t w a r e ; you can r e d i s t r i b u t e i t and / or m o d i f y i t u n d e r t h e t e r m s o f t h e GNU G e n e r a l P u b l i c L i c e n s e a s p u b l i s h e d b y the Free So f tw ar e Foundation ; e i t h e r v e r s i o n 2 of the License , or ( a t your o p t i o n ) any l a t e r v e r s i o n . This program i s d i s t r i b u t e d in the hope t h a t i t w i l l be u s e f u l , b u t WITHOUT ANY WARRANTY ; w i t h o u t e v e n t h e i m p l i e d w a r r a n t y o f MERCHANTABILITY o r FITNESS FOR A PARTICULAR PURPOSE . See th e GNU G e n e r a l P u b l i c L i c e n s e f o r m o r e d e t a i l s . You s h o u l d h a v e r e c e i v e d a c o p y o f t h e GNU G e n e r a l P u b l i c L i c e n s e a l o n g w i t h t h i s program ; i f not , w r i t e t o t h e Free S o f t w a r e F o u n d a t i o n , I n c . , 5 1 F r a n k l i n S t , F i f t h F l o o r , B o s t o n , MA 02110 1301 / #d e f i n e #i n c l u d e #i n c l u d e #i n c l u d e USE MATH DEFINES m a tr ix . h mex . h math . h ( ( x)>=0?( long ) ( ( x ) + 0 . 5 ) : ( long ) ( ( x ) 0 . 5 ) )

USA

#d e f i n e ROUND( x )

#i f n d e f M PI #d e f i n e M PI 3 . 1 4 1 5 9 2 6 5 3 5 8 9 7 9 3 2 3 8 4 6 #e n d i f double { s i n c ( double x )

17

i f ( x == 0 ) return ( 1 . ) ; else return ( s i n ( x ) / x ) ; } double { if { s i m m i c r o p h o n e ( double x , ( mtype== b double || mtype== c double y , || double z , || double a n g l e , char mtype )

mtype== s varphi ,

mtype== h )

strength ,

vartheta ,

alpha ;

// P o l a r P a t t e r n alpha // // B i d i r e c t i o n a l 0 // H y p e r c a r d i o i d 0.25 // C a r d i o i d 0.5 // S u b c a r d i o i d 0.75 // O m n i d i r e c t i o n a l 1 switch ( mtype ) { case b : alpha = 0 ; break ; case h : alpha = 0 . 2 5 ; break ; case c : alpha = 0 . 5 ; break ; case s : alpha = 0 . 7 5 ; break ; }; v a r t h e t a = a c o s ( z / s q r t ( pow ( x , 2 ) + pow ( y , 2 ) + pow ( z , 2 ) ) ) ; varphi = atan2 ( y , x ) ; s t r e n g t h = s i n ( M PI/2 a n g l e [ 1 ] ) s i n ( v a r t h e t a ) + c o s ( M PI/2 a n g l e [ 1 ] ) c o s ( v a r t h e t a ) ; s t r e n g t h = a l p h a + (1 a l p h a ) s t r e n g t h ; return } else { return } } void { if { mexFunction ( i n t ( n r h s == 0 ) m e x P r i n t f ( \n | Room I m p u l s e R e s p o n s e G e n e r a t o r | \ n | | \ n | Computes t h e r e s p o n s e o f an a c o u s t i c s o u r c e t o o n e o r more | \ n | \ n | m i c r o p h o n e s i n a r e v e r b e r a n t room u s i n g t h e i m a g e method [ 1 , 2 ] . | | \ n | Author : d r . i r . Emanuel H a b e t s ( e h a b e t s @ d e r e v e r b e r a t i o n . o r g ) | \ n | | \ n : 2.0.20100920 | \ n | Version | | \ n | C o p y r i g h t (C) 2003 2010 E . A . P . Habets , The N e t h e r l a n d s . | \ n | | \ n | [ 1 ] J . B . A l l e n and D . A . B e r k l e y , | \ n Image method f o r e f f i c i e n t l y s i m u l a t i n g s m a l l room a c o u s t i c s , | \ n | | J o u r n a l A c o u s t i c S o c i e t y o f America , | \ n 65(4) , April 1979 , p 943. | \ n | | | \ n | [ 2 ] P .M. P e t e r s o n , | \ n | Simulating the response o f m u l t i p l e microphones to a s i n g l e | \ n | a c o u s t i c s o u r c e i n a r e v e r b e r a n t room , J o u r n a l A c o u s t i c | \ n | S o c i e t y o f America , 8 0 ( 5 ) , November 1 9 8 6 . | \ n \n\n f u n c t i o n [ h , b e t a h a t ] = r i r g e n e r a t o r ( c , f s , r , s , L , b e t a , nsample , \ n mtype , o r d e r , dim , o r i e n t a t i o n , h p f i l t e r ) ; \ n\n I n p u t p a r a m e t e r s : \ n : s o u n d v e l o c i t y i n m/ s . \ n c fs : s a m p l i n g f r e q u e n c y i n Hz . \ n r : M x 3 a r r a y s p e c i f y i n g t h e ( x , y , z ) c o o r d i n a t e s o f t h e \n r e c e i v e r ( s ) i n m. \ n s : 1 x 3 v e c t o r s p e c i f y i n g t h e ( x , y , z ) c o o r d i n a t e s o f t h e \n s o u r c e i n m. \ n L : 1 x 3 v e c t o r s p e c i f y i n g t h e room d i m e n s i o n s ( x , y , z ) i n m. \ n beta : 1 x 6 v e c t o r s p e c i f y i n g t h e r e f l e c t i o n c o e f f i c i e n t s \n [ b e t a x 1 b e t a x 2 b e t a y 1 b e t a y 2 b e t a z 1 b e t a z 2 ] o r \n b e t a = R e v e r b e r a t i o n Time ( T 60 ) i n s e c o n d s . \ n nsample : number o f s a m p l e s t o c a l c u l a t e , d e f a u l t i s T 60 f s . \ n mtype : [ o m n i d i r e c t i o n a l , s u b c a r d i o i d , c a r d i o i d , h y p e r c a r d i o i d , \ n b i d i r e c t i o n a l ] , d e f a u l t i s o m n i d i r e c t i o n a l . \ n order : r e f l e c t i o n o r d e r , d e f a u l t i s 1, i . e . maximum o r d e r . \ n dim : room d i m e n s i o n ( 2 o r 3 ) , d e f a u l t i s 3 . \ n o r i e n t a t i o n : d i r e c t i o n i n w h i c h t h e m i c r o p h o n e s a r e p o i n t e d , s p e c i f i e d u s i n g \n a z i m u t h and e l e v a t i o n a n g l e s ( i n r a d i a n s ) , d e f a u l t i s [ 0 0 ] . \ n hp filter : u s e f a l s e t o d i s a b l e h i g hp a s s f i l t e r , t h e h i g hp a s s f i l t e r \n i s e n a b l e d by d e f a u l t . \ n\n Output p a r a m e t e r s : \ n h : M x n s a m p l e m a t r i x c o n t a i n i n g t h e c a l c u l a t e d room i m p u l s e \n r e s p o n s e ( s ) . \ n beta hat : I n c a s e a r e v e r b e r a t i o n t i m e i s s p e c i f i e d a s an i n p u t p a r a m e t e r \n t h e c o r r e s p o n d i n g r e f l e c t i o n c o e f f i c i e n t i s r e t u r n e d . \ n\n ) ; return ; nlhs , mxArray p l h s [ ] , int nrhs , c o n s t mxArray p r h s [ ] ) 1; strength ; c o s ( a n g l e [0] v a r p h i )

18

} else { m e x P r i n t f ( Room I m p u l s e R e s p o n s e G e n e r a t o r ( V e r s i o n 2 . 0 . 2 0 1 0 0 9 2 0 ) C o p y r i g h t (C) 2003 2010 E . A . P . Habets , The N e t h e r l a n d s . \ n ) ; } // if Check f o r p r o p e r number o f a r g u m e n t s ( nrhs < 6) mexErrMsgTxt ( E r r o r : T h e r e a r e a t l e a s t s i x i n p u t p a r a m e t e r s i f ( nrhs > 12) mexErrMsgTxt ( E r r o r : Too many i n p u t a r g u m e n t s . ) ; i f ( nlhs > 2) mexErrMsgTxt ( E r r o r : Too many o u t p u t a r g u m e n t s . ) ; Check f o r p r o p e r arguments ( ! ( mxGetN ( p r h s [ 0 ] ) = = 1 ) | | ! m x I s D o u b l e ( p r h s [ 0 ] ) mexErrMsgTxt ( I n v a l i d i n p u t a r g u m e n t s ! ) ; i f ( ! ( mxGetN ( p r h s [ 1 ] ) = = 1 ) | | ! m x I s D o u b l e ( p r h s [ 1 ] ) mexErrMsgTxt ( I n v a l i d i n p u t a r g u m e n t s ! ) ; i f ( ! ( mxGetN ( p r h s [ 2 ] ) = = 3 ) | | ! m x I s D o u b l e ( p r h s [ 2 ] ) mexErrMsgTxt ( I n v a l i d i n p u t a r g u m e n t s ! ) ; i f ( ! ( mxGetN ( p r h s [ 3 ] ) = = 3 ) | | ! m x I s D o u b l e ( p r h s [ 3 ] ) mexErrMsgTxt ( I n v a l i d i n p u t a r g u m e n t s ! ) ; i f ( ! ( mxGetN ( p r h s [ 4 ] ) = = 3 ) | | ! m x I s D o u b l e ( p r h s [ 4 ] ) mexErrMsgTxt ( I n v a l i d i n p u t a r g u m e n t s ! ) ; i f ( ! ( mxGetN ( p r h s [ 5 ] ) = = 6 | | mxGetN ( p r h s [ 5 ] ) = = 1 ) | | mexErrMsgTxt ( I n v a l i d i n p u t a r g u m e n t s ! ) ;

by Emanuel H a b e t s \n

required . );

// if

|| || || || ||

mxIsComplex ( p r h s [ 0 ] ) ) mxIsComplex ( p r h s [ 1 ] ) ) mxIsComplex ( p r h s [ 2 ] ) ) mxIsComplex ( p r h s [ 3 ] ) ) mxIsComplex ( p r h s [ 4 ] ) ) || mxIsComplex ( p r h s [ 5 ] ) )

! mxIsDouble ( p r h s [ 5 ] )

// Load p a r a m e t e r s double c = mxGetScalar ( prhs [ 0 ] ) ; double f s = mxGetScalar ( prhs [ 1 ] ) ; c o n s t double r r = mxGetPr ( p r h s [ 2 ] ) ; int n r o f m i c s = ( i n t ) mxGetM( p r h s [ 2 ] ) ; c o n s t double s s = mxGetPr ( p r h s [ 3 ] ) ; c o n s t double LL = mxGetPr ( p r h s [ 4 ] ) ; c o n s t double b e t a p t r = mxGetPr ( p r h s [ 5 ] ) ; double b e t a = new double [ 6 ] ; int nsamples ; char mtype ; int order ; int dim ; double angle [ 2 ] ; int hp filter ; double TR ; p l h s [ 1 ] = m x C r e a t e D o u b l e M a t r i x ( 1 , 1 , mxREAL ) ; double b e t a h a t = mxGetPr ( p l h s [ 1 ] ) ; beta hat [ 0 ] = 0; // if { Reflection coefficients ( mxGetN ( p r h s [ 5 ] ) = = 1 ) or Reverberation Time ?

double V = LL [ 0 ] LL [ 1 ] LL [ 2 ] ; double S = 2 ( LL [ 0 ] LL [ 2 ] + LL [ 1 ] LL [ 2 ] + LL [ 0 ] LL [ 1 ] ) ; TR = b e t a p t r [ 0 ] ; double a l f a = 24V l o g ( 1 0 . 0 ) / ( c STR ) ; i f ( a l f a > 1) mexErrMsgTxt ( E r r o r : The r e f l e c t i o n c o e f f i c i e n t s c a n n o t be c a l c u l a t e d u s i n g Please room p a r a m e t e r s , i . e . room s i z e and r e v e r b e r a t i o n t i m e . \ n s p e c i f y t h e r e f l e c t i o n c o e f f i c i e n t s o r c h a n g e t h e room p a r a m e t e r s . ) ; b e t a h a t [ 0 ] = s q r t (1 a l f a ) ; f o r ( i n t i =0; i <6; i ++) beta [ i ] = beta hat [ 0 ] ; } else { f o r ( i n t i =0; i <6; i ++) beta [ i ] = b e t a p t r [ i ] ; } // if { H i g h p a s s f i l t e r ( o p t i o n a l ) ( n r h s > 11 && mxIsEmpty ( p r h s [ 1 1 ] ) == f a l s e ) mxGetScalar ( prhs [ 1 1 ] ) ;

the

current

h p f i l t e r = ( int ) } else { h p f i l t e r = 1; } // if {

3D M i c r o p h o n e o r i e n t a t i o n ( o p t i o n a l ) ( n r h s > 10 && mxIsEmpty ( p r h s [ 1 0 ] ) == f a l s e ) c o n s t double o r i e n t a t i o n = mxGetPr ( p r h s [ 1 0 ] ) ; i f ( mxGetN ( p r h s [ 1 0 ] ) == 1 ) { angle [ 0 ] = orientation [ 0 ] ; angle [ 1 ] = 0; } else { angle [ 0 ] = orientation [ 0 ] ; angle [ 1 ] = orientation [ 1 ] ; }

} else { angle [ 0 ] = 0; angle [ 1 ] = 0; } // if { Room D i m e n s i o n ( o p t i o n a l ) ( n r h s > 9 && mxIsEmpty ( p r h s [ 9 ] ) == f a l s e )

19

dim = ( i n t ) m x G e t S c a l a r ( p r h s [ 9 ] ) ; i f ( dim != 2 && dim != 3 ) mexErrMsgTxt ( I n v a l i d i n p u t a r g u m e n t s ! ) ; if { ( dim == 2 ) beta [ 4 ] = 0 ; beta [ 5 ] = 0 ; } } else { dim = 3 ; } // if { Reflection order ( optional ) ( n r h s > 8 && mxIsEmpty ( p r h s [ 8 ] ) == f a l s e ) o r d e r = ( i n t ) mxGetScalar ( prhs [ 8 ] ) ; i f ( o r d e r < 1) mexErrMsgTxt ( I n v a l i d i n p u t a r g u m e n t s ! ) ; } else { o r d e r = 1; } // if { Type o f m i c r o p h o n e ( o p t i o n a l ) ( n r h s > 7 && mxIsEmpty ( p r h s [ 7 ] ) == f a l s e )

mtype = new char [ mxGetN ( p r h s [ 7 ] ) + 1 ] ; m x G e t S t r i n g ( p r h s [ 7 ] , mtype , mxGetN ( p r h s [ 7 ] ) + 1 ) ; } else { mtype = new char [ 1 ] ; mtype [ 0 ] = o ; } // if { Number o f s a m p l e s ( o p t i o n a l ) ( n r h s > 6 && mxIsEmpty ( p r h s [ 6 ] ) == f a l s e )

nsamples = ( i n t ) mxGetScalar ( prhs [ 6 ] ) ; } else { i f ( mxGetN ( p r h s [ 5 ] ) > 1 ) { double V = LL [ 0 ] LL [ 1 ] LL [ 2 ] ; double S = 2 ( LL [ 0 ] LL [ 2 ] + LL [ 1 ] LL [ 2 ] + LL [ 0 ] LL [ 1 ] ) ; double a l p h a = ((1 pow ( b e t a [ 0 ] , 2 ) ) + ( 1 pow ( b e t a [ 1 ] , 2 ) ) ) LL [ 0 ] LL [ 2 ] + ((1 pow ( b e t a [ 2 ] , 2 ) ) + ( 1 pow ( b e t a [ 3 ] , 2 ) ) ) LL [ 1 ] LL [ 2 ] + ((1 pow ( b e t a [ 4 ] , 2 ) ) + ( 1 pow ( b e t a [ 5 ] , 2 ) ) ) LL [ 0 ] LL [ 1 ] ; TR = 24 l o g ( 1 0 . 0 ) V/ ( c a l p h a ) ; i f (TR < 0 . 1 2 8 ) TR = 0 . 1 2 8 ; } n s a m p l e s = ( i n t ) (TR f s ) ; } // C r e a t e o u t p u t v e c t o r p l h s [ 0 ] = mxCreateDoubleMatrix ( n r o f m i c s , double imp = mxGetPr ( p l h s [ 0 ] ) ; // Temporary c o n s t double c o n s t double c o n s t double c o n s t double c o n s t double double double // Temporary c o n s t double const i n t c o n s t double double double double double double double double double double double int int int int int v a r i a b l e s and c o n s t a n t s W = 2 M PI 1 0 0 / f s ; R1 = exp( W) ; B1 = 2R1 c o s (W) ; B2 = R1 R1 ; A1 = (1+R1 ) ; X0 ; Y = new double [ 3 ] ;

n s a m p l e s , mxREAL ) ;

( h i g h p a s s

filter )

v a r i a b l e s a n d c o n s t a n t s ( i m a g e m e t h o d ) Fc = 1 ; Tw = 2 ROUND( 0 . 0 0 4 f s ) ; cTs = c / f s ; h a n n i n g w i n d o w = new double [ Tw+ 1 ] ; LPI = new double [ Tw+ 1 ] ; r = new double [ 3 ] ; s = new double [ 3 ] ; L = new double [ 3 ] ; hu [ 6 ] ; refl [3]; dist ; ll ; strength ; pos , f d i s t ; n1 , n2 , n3 ; q, j , k; mx , my , mz ; n; s [ 1 ] = s s [ 1 ] / cTs ; L [ 1 ] = LL [ 1 ] / cTs ; s [ 2 ] = s s [ 2 ] / cTs ; L [ 2 ] = LL [ 2 ] / cTs ;

s [ 0 ] = s s [ 0 ] / cTs ; L [ 0 ] = LL [ 0 ] / cTs ;

// Hanning window f o r ( n = 0 ; n < Tw+1 ; n++) { Tw/ 2 ) /Tw ) ) ; h a n n i n g w i n d o w [ n ] = 0 . 5 ( 1 + c o s ( 2 M PI ( n+ } f o r ( i n t m i c n r = 0 ; m i c n r < n r o f m i c s ; m i c n r ++) { // [ x 1 x 2 . . . x N y 1 y 2 . . . y N z 1 z 2 . . . z N ]

20

r [ 0 ] = r r [ m i c n r + 0 n r o f m i c s ] r [ 1 ] = r r [ m i c n r + 1 n r o f m i c s ] r [ 2 ] = r r [ m i c n r + 2 n r o f m i c s ] n1 = ( i n t ) n2 = ( i n t ) n3 = ( i n t )

/ cTs ; / cTs ; / cTs ;

c e i l ( nsamples /(2L [ 0 ] ) ) ; c e i l ( nsamples /(2L [ 1 ] ) ) ; c e i l ( nsamples /(2L [ 2 ] ) ) ; response ; mx++)

// G e n e r a t e room i m p u l s e f o r (mx = n1 ; mx <= n1 { hu [ 0 ] = 2mxL [ 0 ] ;

f o r (my = n2 ; my <= n2 { hu [ 1 ] = 2myL [ 1 ] ;

; my++)

f o r ( mz = n3 ; mz <= n3 { hu [ 2 ] = 2mzL [ 2 ] ;

; mz++)

f o r ( q = 0 ; q <= 1 ; q++) { hu [ 3 ] = (1 2 q ) s [ 0 ] r [ 0 ] + hu [ 0 ] ; r e f l [ 0 ] = pow ( b e t a [ 0 ] , a b s (mx q ) ) pow ( b e t a [ 1 ] ,

a b s (mx ) ) ;

f o r ( j = 0 ; j <= 1 ; j ++) { hu [ 4 ] = (1 2 j ) s [ 1 ] r [ 1 ] + hu [ 1 ] ; r e f l [ 1 ] = pow ( b e t a [ 2 ] , a b s (my j ) ) pow ( b e t a [ 3 ] , f o r ( k = 0 ; k <= 1 ; k++) { hu [ 5 ] = (1 2 k ) s [ 2 ] r [ 2 ] + hu [ 2 ] ; r e f l [ 2 ] = pow ( b e t a [ 4 ] , a b s ( mzk ) ) pow ( b e t a [ 5 ] , d i s t = s q r t ( pow ( hu [ 3 ] , if { 2 ) + pow ( hu [ 4 ] ,

a b s (my ) ) ;

a b s ( mz ) ) ; 2));

2 ) + pow ( hu [ 5 ] , ||

( a b s ( 2 mx q)+ a b s ( 2 my j )+ a b s ( 2 mzk ) <= o r d e r f d i s t = ( int ) ( f d i s t < nsamples ) floor ( dist );

o r d e r == 1)

if {

s t r e n g t h = s i m m i c r o p h o n e ( hu [ 3 ] , hu [ 4 ] , hu [ 5 ] , r e f l [ 0 ] r e f l [ 1 ] r e f l [ 2 ] / ( 4 M PI d i s t cTs ) ; f o r ( n = 0 ; n < Tw+1 ; n++) LPI [ n ] = h a n n i n g w i n d o w [ n ] Fc

angle ,

mtype [ 0 ] )

s i n c ( M PIFc ( n( d i s t f d i s t ) (Tw / 2 ) ) ) ;

p o s = f d i s t (Tw/ 2 ) ; f o r ( n = 0 ; n < Tw+1; n++) i f ( p o s+n >= 0 && p o s+n < n s a m p l e s ) imp [ m i c n r + n r o f m i c s ( p o s+n ) ] += s t r e n g t h } } } } } } } } // if { O r i g i n a l h i g h p a s s ( h p f i l t e r == 1 ) filter as proposed by Allen and Berkley .

LPI [ n ] ;

f o r ( i n t i d x = 0 ; i d x < 3 ; i d x ++) {Y [ i d x ] = 0 ; } f o r ( i n t i d x = 0 ; i d x < n s a m p l e s ; i d x ++) { X0 = imp [ m i c n r+n r o f m i c s i d x ] ; Y[ 2 ] = Y[ 1 ] ; Y[ 1 ] = Y[ 0 ] ; Y [ 0 ] = B1Y [ 1 ] + B2Y [ 2 ] + X0 ; imp [ m i c n r+n r o f m i c s i d x ] = Y [ 0 ] + A1Y [ 1 ] + R1Y [ 2 ] ; } } } }

References
[1] J. Allen and D. Berkley, Image Method for Eciently Simulating Small Room Acoustics, Journal of the Acoustical Society of America, vol. 65, no. 4, pp. 943950, 1979. [2] H. Kuttru, Room Acoustics, 4th ed. London: Spon Press, 2000.

[3] L. Savioja, J. Huopaniemi, T. Lokki, and R. Vnnen, Creating interactive virtual aa a acoustic environments, Journal of the Audio Engineering Society, vol. 47, no. 9, pp. 675705, 1999. [4] A. Kulowski, Algorithmic representation of the ray tracing technique, Applied Acoustics, vol. 18, no. 6, pp. 449469, 1985.

21

[5] M. Kleiner, B. Dalenbck, and P. Svensson, Auralization - an overview, Journal of a the Audio Engineering Society, vol. 41, no. 11, pp. 861875, Nov. 1993. [6] A. Pietrzyk, Computer modeling of the sound eld in small rooms, in Proc. of the 15th AES Int. Conf. on Audio, Acoustics & Small Spaces, vol. 2, Copenhagen, Denmark, Oct. 1998, pp. 2431. [7] D. Botteldoore, Finite-dierence time-domain simulation of low-frequency room acoustic problems, Journal of the Acoustical Society of America, vol. 98, no. 6, pp. 3302 3308, 1995. [8] L. Savioja, J. Backman, A. Jrvinen, and T. Takala, Waveguide mesh method for a low-frequency simulation of room acoustics, in Proc. of the 15th Int. Congr. Acoust. (ICA95), vol. 2, Trondheim, Norway, Jun. 1995, pp. 14. [9] M. Crocker, Handbook of Acoustics. Wiley-Interscience, 1998.

[10] P. Peterson, Simulating the response of multiple microphones to a single acoustic source in a reverberant room, Journal of the Acoustical Society of America, vol. 80, no. 5, pp. 15271529, Nov. 1986. [11] A. Pierce, Acoustics: An Introduction to Its Physical Principles and Applications. Acoustical Society of America, 1991.