EP2360940A1

EP2360940A1 - Steerable microphone array system with a first order directional pattern

Info

Publication number: EP2360940A1
Application number: EP10151106A
Authority: EP
Inventors: Bart De Schuymer; Henk Brouckxon
Original assignee: Televic NV
Current assignee: Televic NV
Priority date: 2010-01-19
Filing date: 2010-01-19
Publication date: 2011-08-24
Also published as: BE1019734A3

Abstract

The present invention is related to a microphone array system (1) for producing an output signal with a first order directional pattern comprising
- a first (2) and a second (4) omnidirectional microphone located at a distance from each other which is smaller than a minimum acoustic wavelength defined by a given audio frequency range of operation,
- a first filtering means (6) for filtering a signal received by said first omnidirectional microphone and a second filtering means (8) for filtering a signal received by said second omnidirectional microphone, said first and second filtering means each having a frequency response that takes into account said frequency dependent sensitivity of the corresponding microphone,
- summing means (9) for summing the signals output by said first and second filtering means, so that the resulting summed signal has said first order directional pattern.

Description

Field of the Invention

The present invention generally relates to the field of microphone arrays as used, for example, in conference systems.

Background of the Invention

It is well known in the art how to approximate a so-called first order polar pattern by using an array of two or more microphones. The microphones are pressure-sensitive microphones each sensing the acoustic pressure at a single point. The polar pattern is formed as the difference in pressure between two points in space. It constitutes an indication of the sensitivity to sounds arriving at different angles about its central axis. Polar patterns represent the location of points that produce the same signal level output in the microphone if a given sound pressure level is generated from that point. A polar pattern is an indication of the microphone's directionality.
A first order differential microphone generally has a polar pattern given by R(θ) = A + (1 - A) cos θ, with 0 ≤ A ≤ 1 and θ the angle of inclination (azimuth), so that the response is normalized to have a maximum value of 1 at θ = 0. The expression for R(θ) is the parametric expression for the 'limaçon of Pascal' algebraic curve, which is well known to those of skill in the art. The two terms in the above equation can be seen to be the sum of an omnidirectional sensor (i.e., the first-term) and a first-order dipole sensor (i.e., the second term), which is the general form of the first-order array. Fig.1 represents polar plots of three well-known members of the limaçon family, namely the dipole (A = 0), the cardioid (A = 0.5) and the hyper-cardioid (A = 0.25). It is to be noted that these polar plots always use the absolute value of R(θ).
Patent document EP 0869697 B1 relates to a steerable and variable first-order differential microphone array. The invention is also described in the paper "A steerable and variable first-order differential microphone array" (G.W. Elko et al., ICASSP'97, pp. 223-226). In the proposed method a polar pattern of type A + B cos θ is being approximated (with A = 1-B). This pattern can be seen as the sum of an omnidirectional microphone (A) and a bi-directional (also called figure-of-eight) microphone (B cosθ). In fact, the first commercial cardioid and poly-directional microphones were made by summing the signals of these two types of microphones. To obtain the desired polar pattern a delay-and-sum approach is followed. A microphone array setup used is shown in Fig.2. When simply subtracting the second microphone's input from the first one's input, the following polar pattern is obtained (with k denoting the acoustic wave number) : $R (θ) = e^{- j \frac{kd}{2} \cos θ} - e^{j \frac{kd}{2} \cos θ},$

or, equivalently, $R (θ) = - 2 j \sin (\frac{kd}{2} \cos θ) .$

Noting that sin x ≈ x for small values of x, one obtains $R (θ) \approx - jkd \cos θ,$

for kd << π. The above equation is readily recognized as a 90° phase shifted figure-of-eight polar pattern. The scaled addition of an omnidirectional 90° phase-shifted polar pattern with the above approximation of a dipole polar pattern enables the construction of an approximation of any first-order (limaçon) polar pattern. This phase-shifted omnidirectional pattern is achieved by adding the scaled output of two oppositely directed phase-shifted cardioid approximations. A cardioid polar pattern can be approximated by adding the output of the first microphone to a delayed (by kd seconds) output of the second microphone. The oppositely directed polar pattern is obtained by delaying the first microphone instead of the second. Summing these opposite approximations of cardioid polar patterns (scaled by 1/2) gives: $R (θ) = \frac{1}{2} ((- e^{j \frac{kd}{2}} + e^{- j \frac{kd}{2}}) e^{- j \frac{kd}{2} \cos θ} + (- e^{j \frac{kd}{2}} + e^{- j \frac{kd}{2}}) e^{j \frac{kd}{2} \cos θ}),$

which simplifies to $R (θ) = - j \sin (\frac{kd}{2} (1 + \cos θ)) - j \sin (\frac{kd}{2} (1 - \cos θ)),$

and finally $R (θ) = - 2 j \sin (\frac{kd}{2}) \cos (kd) (\frac{}{2} cosθ) .$

For kd << π the above equation approximates an omnidirectional polar pattern, with the same phase shift as the obtained figure-of-eight polar pattern approximation. All polar patterns are obtainable with the delay-and-sum technique, as the filter coefficients are frequency-independent. The proposed solution relies on the application of analogue filters. When plotting the magnitude response for θ = 0, a high-pass characteristic is readily observed for the applied delay-and-sum. Undoing this high-pass behaviour requires an additional filter and also amplifies the low-frequency noise. A further drawback of the delay-and-sum method is that unless the filter is implemented in an analogue circuit, a fractional delay is needed, e.g. about 2.09913 samples for d = 0,015 m and a sampling frequency of 48kHz. The accuracy of the fractional delay depends on the used filter length. Obtaining a good fractional delay accuracy therefore implies adding extra delay. Yet another drawback of the delay-and-sum approach is the undesirable effect of attenuating the signal when the microphone distance d is decreased. This is illustrated in Fig.3. On the left side the polar patterns are shown for the delay-and-sum method with d = 0.015m, on the right side the result is shown for d = 0.005m. It is readily verified that, while the shape of the polar pattern improves, the gain decreases when d decreases.

Aims of the invention

The present invention aims to provide a microphone array system capable of producing an output signal with a first order directional pattern wherein the drawbacks of the prior art are overcome. The invention further aims to provide a method for generating an output signal with a first order directional pattern.

Summary

The present invention relates in a first aspect to a microphone array system capable of producing an output signal having a first order directional pattern. The microphone array system comprises a first and a second omnidirectional microphone. These two microphones are located at a distance from each other which is smaller than a minimum acoustic wavelength defined by the desired audio frequency range of operation. The microphone array system comprises a first filter for filtering a signal received by the first omnidirectional microphone and a second filter for filtering a signal received by the second omnidirectional microphone. The first filter has a first frequency response. This filter yields a first filtered output signal. The second filter has a second frequency response and yields a second filtered output signal. The first filter and second filter are devised such that they have a frequency response that takes into account the frequency response of the first and second omnidirectional microphone. The microphone array system further comprises summing means for summing the output signals of the first and second filter. The resulting combined signal then has the desired first order directional pattern.
The proposed solution indeed achieves the goal. Using only two microphones good directional patterns are obtainable over a wide frequency range, thus enabling cheap and small solutions. The followed approach can be seen as a filter-and-sum method, whereas prior art solutions rather use a delay-and-sum method. The filter-and-sum method displays low-pass behaviour, whereas the delay-and-sum rather has a high-pass characteristic, as mentioned previously.
In a preferred embodiment the first and second filtering means are implemented as finite-impulse response filters. The proposed solution indeed is advantageously carried out in digital implementation. An implementation of the delay-and-sum method would need a fractional delay. The accuracy of the fractional delay depends on the used filter length. Obtaining sufficient fractional delay accuracy therefore implies adding extra delay. This problem is avoided by applying the filter-and-sum approach of the present invention.
In a digital implementation of the microphone array system preferably conversion means are provided for converting the respective signals received by the first and second omnidirectional microphone into corresponding digital output signals. In an alternative embodiment digital microphones of sufficient quality can be used, eliminating the analogue-to-digital conversion.
In an advantageous embodiment the microphone array system further comprises means for performing block processing for determining the signals output by the first and second filtering means. This enables faster processing, using the overlap-add technique when performing the filtering operations.
In a preferred embodiment the first and second filtering means are adaptable, so that the parameters A and B can be set of the first order directional pattern of type A+Bcos θ.
In one embodiment of the invention the microphone array system further comprises a third omnidirectional microphone and a third filter for filtering the signal received by the third omnidirectional microphone. This third microphone is positioned at a distance from the first or the second microphone which is smaller than the minimum acoustic wavelength defined by the audio frequency range of operation. It is preferably so placed that the set of three microphones are the vertices of a substantially right and isosceles triangle. The third filter has a frequency response determined by taking into account the frequency response of the third omnidirectional microphone. The summing means is so organised that the filtered signal output by the third filter can be added to the signals output by the first and second filter.
In a preferred embodiment the microphone array system comprises a third and a fourth omnidirectional microphone. The third and fourth omnidirectional microphone are located at a distance from each other which is smaller than the minimum acoustic wavelength defined by the given audio frequency range of operation. The third and fourth microphone are so positioned that they form an array with an axis substantially orthogonal with the axis of the array formed by said first and second microphone. Preferably the centres of the four microphones form the corners of substantially a square. The third and the fourth filter have a third and a fourth frequency response, respectively. These frequency responses take into account the frequency responses of the third and fourth omnidirectional microphone. The summing means in this embodiment is preferably arranged for summing the signals output by the third and fourth filtering means and for combining the resulting output signal with the resulting output signal output by said first and second filtering means. Alternatively, separate summing means for the signals output by the third and fourth filtering means are provided as well as suitable combining means for combining the sum signals of the first and second filter and the third and fourth filter, respectively. This configuration allows steering the listening direction over 360 degrees.
In another preferred embodiment the microphone array system further comprises storage means for storing filter coefficient values for a plurality of steering angles. In a setup with two microphones steering is only possible in the two directions of the microphone axis. With more microphones available any listening direction in the plane formed by the microphones can be steered.
The invention also relates to a speaker unit of a conference system comprising a microphone array system as previously described.
Advantageously, the speaker unit is arranged for being built in into a table. In this way, the microphone array is completely non-intrusive and practically invisible to the speaker: the flat surface of the table is preserved. This is possible due to the use of omnidirectional microphones. In another advantageous embodiment the microphone array system can be built-in in a table, in a fold-away manner, such that it can be lifted up when needed, thereby reducing the elevation angle.
The invention also relates to a conference system comprising a plurality of said speaker units.
In another aspect the invention relates to a method for generating a microphone array output signal having a first order directional pattern, comprising the steps of

providing a first and a second microphone located at a distance from each other which is smaller than a minimum acoustic wavelength defined by a given audio frequency range of operation,
applying a signal received by the first microphone to a first filtering means and a signal received by the second microphone to a second filtering means, whereby the first and second filtering means have a first and second frequency response, respectively, whereby this first and second frequency response are determined taking into account the frequency response of the third and fourth omnidirectional microphone,
summing the signals output by the first and second filtering means, yielding the microphone array output signal with first order directional pattern.

In a most preferred embodiment the first and second microphone are configured as an end-fire array, whereby the listening angle is parallel to the microphone axis.
The above method can also be applied with slight adaptations when a system with three or four microphones is used.

Brief Description of the Drawings

Fig. 1 illustrates some well-known polar patterns.
Fig. 2 represents a first-order microphone array setup.
Fig. 3 illustrates the microphone distance dependence problem of the delay-and-sum method.
Fig. 4 represents a two microphone setup
Fig. 5 represents the filter-and-sum method of the present invention.
Fig. 6 illustrates directional patterns obtained with a two microphone setup.
Fig. 7 represents a steerable first-order method configuration.
Fig. 8 represents a rotated cardioid dissection.
Fig. 9 represents directional patterns obtained with a four microphone setup.
Fig. 10 represents improvements by using 4 microphones instead of 2.
Fig. 11 represents a conference system setup.
Fig. 12 represents some two-dimensional polar plots for a few elevation angles.
Fig. 13 represents a three microphone setup.
Fig. 14 represents an end-fire array with three microphones.

Detailed Description of the Invention

In a first aspect the present invention relates to a microphone array system for producing an output signal having a first order directional pattern. As opposed to EP0869697 an approach is followed wherein a filter-and-sum method is applied for obtaining an approximation of a first-order polar pattern. Good directional patterns can so be achieved over a wide frequency range using only two (omnidirectional) microphones, enabling cheap and small solutions.
The first-order microphone array setup as shown in Fig.4 is considered. Note that in the calculations below the reference point is in the middle between the two microphones. The frequency responses of the first and second omnidirectional microphone can be written as $S_{1} (ω θ) = \exp (- j \frac{ω . d}{2. c} . \cos (θ)) \approx (1 - j \frac{ω . d}{2. c} . \cos (θ))$
$S_{2} (ω θ) = \exp (+ j \frac{ω . d}{2. c} . \cos (θ)) \approx (1 + j \frac{ω . d}{2. c} . \cos (θ))$

whereby d is the distance between the two microphones and ω the angular frequency. The ω/c can be written as ω/c = 2πflc = 2π/λ = k, with λ representing the wavelength of the sound signal, f the acoustic frequency, c the speed of sound in air (approximately 343 m/s) and k the acoustic wave number. θ denotes the angle of inclination. In the approximation step above, the first two terms of the Taylor polynomial for the exponential function are withheld. The invention proposes to apply a filter H₁ (ω) to S₁ (ω,θ) and a filter H₂ (ω) to S₂ (ω,θ), and to sum the filtered signals, as follows : $\begin{array}{l} S (ω θ) & = H_{1} (ω) . S_{1} (ω θ) + H_{2} (ω) . S_{2} (ω θ) \\ \approx [H_{1} (ω) (1 - j \frac{ω . d}{2. c} . \cos (θ)) + H_{2} (ω) (1 + j \frac{ω . d}{2. c} . \cos (θ))] \\ = [(H_{1} (ω) + H_{2} (ω)) + j (- H_{1} (ω) + H_{2} (ω)) . \frac{ω . d}{2. c} . \cos (θ)] \\ = A + B . \cos (θ) \end{array}$

So the resulting values of A and B are : ${\begin{matrix} A = H_{1} (ω) + H_{2} (ω) \\ B = j (- H_{1} (ω) + H_{2} (ω)) . \frac{ω . d}{2. c} \end{matrix}$

For a given value of A and B one gets the following filter design : ${\begin{matrix} H_{1} (ω) = \frac{A}{2} + j \frac{c}{ωd} B \\ H \\ _{2} (ω) = \frac{A}{2} - j \frac{c}{ωd} B \end{matrix}$

In case the reference point is not taken as the middle between the two microphones the expressions for the filter characteristics can easily be derived in a similar way as shown above. Fig.5 illustrates the approach followed in the filter-and-sum method. It is readily verified that H₁ (ω) and H₂ (ω) depend on the frequency. Unlike the approach followed in "Directional patterns obtained from two or three microphones" (S. Thompson, Knowles Electronics, September 29, 2000), the solution according to the invention advantageously adapts to this frequency dependency.
In an advantageous embodiment a digital implementation with finite impulse response filters is applied.
A transform domain representation is constructed of two FIR filters S ₁ and S _-1 as follows. For a sampling rate of F Hz and a filter length of N taps (where N is even), the values S₁ [n] resp. S _-1[n] for the n^th (0 ≤ n < N) frequency bin of the N point FFT of filters H₁ and H₂ , respectively, are given by: $\begin{matrix} H_{1} [n] = \frac{A}{2} + jB \frac{N}{n} \frac{c}{2 πdF}, \\ H_{2} [n] = \frac{A}{2} - jB \frac{N}{n} \frac{c}{2 πdF}, \end{matrix} for 0 < n < N / 2,$

and $\begin{matrix} H_{1} [n] = \frac{A}{2} + jB \frac{N}{n - N} \frac{c}{2 πdF}, \\ H_{2} [n] = \frac{A}{2} - jB \frac{N}{n - N} \frac{c}{2 πdF}, \end{matrix} for N / 2 \leq n < N$

whereby H ₁[0] = 0.5 and H ₂[0]=0.5. A sampling frequency of e.g. 48 kHz and filter length of e.g. N = 1024 can be used. If time-domain filtering is more desirable, the time-domain representation of the filters is readily obtained using an inverse DFT algorithm.
As explained, the desired polar pattern response is obtained by filtering the input of microphone m₁ and m₂ with the FIR filter corresponding to H₁ and H₂ , respectively, and summing the filter outputs. This filtering can e.g. be done in the transform domain, using overlap-add to obtain the correct time domain result. To improve the performance, block processing in the transform domain can be used. Also, to smoothen the filter response, the FIR filter coefficients can be weighted by a window (e.g. the Hamming window) after which they are again transformed to the transform domain.
As also disclosed in EP08696971 , an additional short FIR filter can optionally be provided to compensate for non-idealities of the microphones. This allows the use of common, cheaper microphones. This additional filter can in one embodiment be combined in a filter that corresponds to a cascade of the short filter and the filter H as described above.
The distance between the centres of both microphones in a two microphone setup should be at most about 1.5 cm, because the performance for higher frequencies depends on the microphone distance. A distance of 1.5 cm allows the use of electret microphones with a diameter up to 1.5 cm. The microphones are advantageously placed in end-fire configuration. This means that the listening angle is parallel to the microphone axis. Steering is only possible in the two endfire directions, as shown in Fig.6.
The polar plot of the ideal first-order polar patterns aimed at an angle α can be obtained using the formula $R_{α} (θ) = A + B \cos (α) \cos (θ) + B \sin (α) \sin (θ),$

where (θ,abs(R _α(θ))) are the polar coordinates of the corresponding polar pattern. The well-known cardioid and hypercardioid patterns are obtained by setting A=B=0.5 and A=0.25 and B=0.75, respectively. Note that setting α = 0 results in the limaçon family.
These first-order polar patterns can be approximated by the combination of two orthogonal end-fire arrays with each two microphones, see Fig.6. The configuration consists of four omnidirectional microphones placed on the vertices of a square. The length of the square's diagonal is preferably 1.5 cm or smaller. The two microphone arrays are formed by the two pairs of microphones at the opposite vertices of the square, as indicated by dashed lines in Fig.7.
The rotated cardioid polar pattern for example can be obtained by two orthogonal figure-of-eight polar patterns and one omnidirectional pattern, see Fig.8. The two figure-of-eight patterns add up in such a way that a rotated figure-of-eight pattern is obtained (note that the polar patterns in Fig.8 only give magnitude information, no phase information). By adding this pattern to an omnidirectional pattern (with magnitude 0.5) the desired rotated (in this case 45°) cardioid polar pattern is obtained.
Using the approach according to the invention as described above a steerable microphone array can be implemented. For a polar pattern with parameters A and B rotated α degrees, the first microphone array is of type R ₁(θ)=A/2+B cos α and the second of type R ₂(θ) = A/2 + B sin α. Note that, unlike the method described in EP0869697 , there is no phase shift between the outputs of both microphone arrays, which simplifies the implementation.
With more mathematical detail the filter characteristic derivation can be described as follows. Four microphones m₁ , m₂, m₃ and m₄ are considered (see Fig.7). The centre point is taken as the reference point in the calculations below. Microphone pairs (m₁, m₂ ) and (m₃, m₄ ) can be used to obtain an approximation of the following characteristics, respectively: $(m_{1} m_{2}) \to S_{1, 2} (ω θ) = A_{1, 2} + B_{1, 2} . \cos (θ)$
$(m_{3} m_{4}) \to S_{3, 4} (ω θ) = A_{3, 4} + B_{3, 4} . \cos (θ - \frac{π}{2}) = A_{3, 4} + B_{3, 4} . \sin (θ)$

By using the identity cos(θ + ϕ) = cos(θ)cos(ϕ) - sin(θ)sin(ϕ), a first-order gradient characteristic rotated ϕ degrees can be written as a combination of these characteristics: $\begin{array}{l} S (ω φ θ) & = A + B . \cos (θ + φ) \\ = A + B . (\cos (θ) \cos (φ) - \sin (θ) \sin (φ)) \\ = (\frac{A}{2} + [B \cos (φ)] \cos (θ)) + (\frac{A}{2} - [B \sin (φ)] \sin (θ)) \end{array}$

Resulting in: ${\begin{matrix} A_{1, 2} = A_{3, 4} = A / 2 \\ B_{1, 2} = B \cos (φ) \\ B_{3, 4} = - B \sin (φ) \end{matrix}$

and ${\begin{matrix} H_{1} (ω φ) = \frac{A}{4} + j \frac{c}{ωd} B \cos (φ) \\ H_{2} (ω φ) = \frac{A}{4} - j \frac{c}{ωd} B \cos (φ) \\ H_{3} (ω φ) = \frac{A}{4} - j \frac{c}{ωd} B \sin (φ) \\ H \\ _{4} (ω φ) = \frac{A}{4} + j \frac{c}{ωd} B \sin (φ) \end{matrix}$
The filter coefficients for different steering angles can be efficiently calculated in real-time. Alternatively, the filter coefficients can be calculated off-line and e.g. stored in external memory, so that no computation power is lost for obtaining the filter coefficients.
In a four microphone setup the centres of the four microphones are advantageously placed on the corners of substantially a square. Similar to the two-microphone setup, the square's diagonal length should not exceed 1.5 cm. The setup is illustrated in Fig.9, where two possible (ideal) polar patterns are given.
Even in the case where the steering is only done in one fixed direction parallel to a microphone array axis, it can be advantageous to use the four microphone setup instead of the two microphone setup. This is illustrated in Fig.10, where the 3-dimensional polar pattern is plotted from two viewing angles for a frequency of 8 kHz and a microphone distance of 1.5 cm. The lower two plots correspond with the case where all four microphones are used while the upper two plots correspond with the case where only two microphones are used. In the upper figures a backward lobe is visible which is undesired when approximating a cardioid pattern. It can be concluded that the addition of two microphones on a perpendicular axis is advantageous as the polar pattern for the four microphone setup clearly resembles the ideal cardioid pattern more closely.
With the four microphone setup it is possible to steer the listening direction 360 degrees (without physically altering the microphone array setup). In the two-microphone setup, it is only possible to steer to the front or the back of the microphone array.
Ideally, the plane in which the microphones are placed points (vertically) to the speaker direction, because the directivity is maximized in this plane. However, this is not essential and some deviation from this optimum is allowed. The elevation angle ϕ should be kept as small as possible. A setup in which the microphones are placed in the horizontal plane, for example, should already have sufficient directivity in some situations. This is e.g. the case when the microphones are integrated in a table and the speaker is sitting at the same table (conference system setup). Fig.11 illustrates this setup, in which the elevation angle ϕ is clearly indicated. The 2-dimensional polar plots for a few elevation angles are given in Fig.12. The figures indicate that an elevation angle of 0° to 45° is acceptable.
In one embodiment the microphone array system comprises three microphones. See Fig.13. Such a three microphone array can be split into two subarray pairs $(m_{1}^{*} m_{2})$
and $(m_{1}^{* *} m_{3}),$
with $m_{1}^{*}$
and $m_{1}^{* *}$
physically corresponding to microphone m ₁ : $S (ω φ θ) = (\frac{A}{2} + [B \cos (φ)] \cos (θ)) + (\frac{A}{2} - [B \sin (φ)] \sin (θ))$

and $\begin{matrix} {\begin{matrix} \begin{matrix} A_{1, 2} = A_{1, 3} = A / 2 \\ B_{1, 2} = B \cos (φ) \\ B_{1, 3} = B \sin (φ) \end{matrix} \to {\begin{matrix} H_{1}^{*} (ω φ) = \frac{A}{2} + j \frac{c}{ωd} B \cos (φ) \\ H_{2} (ω φ) = - j \frac{c}{ωd} B \cos (φ) \\ H_{1}^{* *} (ω φ) = \frac{A}{2} + j \frac{c}{ωd} B \sin (φ) \\ H_{3} (ω φ) = - j \frac{c}{ωd} B \sin (φ) \end{matrix} \end{matrix} \\ \to {\begin{matrix} H_{1} (ω φ) = A + j \frac{c}{ωd} B \cos (φ) + j \frac{c}{ωd} B \sin (φ) \\ H \\ _{2} (ω φ) = - j \frac{c}{ωd} B \cos (φ) \\ H_{3} (ω φ) = - j \frac{c}{ωd} B \sin (φ) \end{matrix} \end{matrix}$

As shown in Fig.13, the three microphones are preferably so positioned that they (i.e. their centres) substantially form the vertices of a right and isosceles triangle. The summing means is arranged for also adding the signal output by the third filter. The three microphone solution can advantageously be used as a fall back solution in case one microphone of a 4 microphone system is defunct.
The invention also relates to a speaker unit of a conference system comprising a microphone array system as described. In a conferencing system the participants are seated at tables. The preferred placement of the microphone array system according to this invention is therefore either on the table or mounted inside the table. The selection of the desired steering direction can either be done automatically or manually. Automatic steering can be achieved by determining the direction of the speaker. This can e.g. be done by using a voice activity detector and measuring the root mean square amplitude of the output signal for various steering angles. Another possibility is using a mechanism triggered by a participant or director. This can e.g. be done by pressing a button. Each button then corresponds to a fixed and a priori known steering angle. Some advantageous possible setups are discussed in the following paragraphs.
In a parallel setup one or two participants can utilize the microphone array. Both the two and four microphone setup can be used. Each participant is seated so that the microphone axis is pointing in their direction. When two participants are supported, they are located on opposite sides of the microphone array. When only one participant is talking, a cardioid or hyper-cardioid pattern is used to steer to the participant. When both participants are talking, a dipole pattern can be used to steer to both participants.
In a parallel and perpendicular setup four microphones are required. This setup requires the 4 microphone setup. Up to four participants are supported, but their possible positions are restricted so that all listening angles are parallel to the axis formed by one of the two microphone pairs. In this way the polar patterns are optimized. The participants can be seated each on a side of a square table or alternatively at a round table at 90° angles. Again, a dipole pattern or a combination of a cardioid and dipole pattern can be used if more than one participant is talking.
A round-table setup also requires the four microphone setup. The participants are ideally placed at a round table, although different setups with e.g. a rectangular table are also possible. Now, the ideal situation where each steering direction is parallel to an axis formed by a microphone pair is no longer required. Although the obtained polar patterns will in general degrade somewhat, more flexibility in the placement of the participants is obtained.
The above description explains how a signal with a first order directional pattern can be derived applying a filter-and-sum approach. This approach can be applied for obtaining a higher order directional pattern. For example, if one wishes to generate a signal with a second order pattern, the filters can be determined by taking into account one more term (i.e. the second order term) of the Taylor polynomial of the exponential function. The second order polar pattern that is approximated is then given by $A + B \cos θ + {C \cos}^{2} θ,$

with A + B + C = 1.
In this case, an end-fire array with three, equally spaced, microphones is required (see Fig. 14). The filter design is then characterised by ${\begin{cases} H_{1} (ω) = A + 8 {(\frac{c}{ωd})}^{2} C \\ H \\ _{2} (ω) = - 4 {(\frac{c}{ω d})}^{2} C - j \frac{c}{ωd} B \\ H_{3} (ω) = - 4 {(\frac{c}{ω d})}^{2} C + j \frac{c}{ωd} B \end{cases}$

Note that this microphone array can only be steered in a direction parallel to the microphone axis.
Although the present invention has been illustrated by reference to specific embodiments, it will be apparent to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied with various changes and modifications without departing from the scope thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. In other words, it is contemplated to cover any and all modifications, variations or equivalents that fall within the scope of the basic underlying principles and whose essential attributes are claimed in this patent application. It will furthermore be understood by the reader of this patent application that the words "comprising" or "comprise" do not exclude other elements or steps, that the words "a" or "an" do not exclude a plurality, and that a single element, such as a computer system, a processor, or another integrated unit may fulfil the functions of several means recited in the claims. Any reference signs in the claims shall not be construed as limiting the respective claims concerned. The terms "first", "second", third", "a", "b", "c", and the like, when used in the description or in the claims are introduced to distinguish between similar elements or steps and are not necessarily describing a sequential or chronological order. Similarly, the terms "top", "bottom", "over", "under", and the like are introduced for descriptive purposes and not necessarily to denote relative positions. It is to be understood that the terms so used are interchangeable under appropriate circumstances and embodiments of the invention are capable of operating according to the present invention in other sequences, or in orientations different from the one(s) described or illustrated above.

Claims

A microphone array system (1) for producing an output signal with a first order directional pattern comprising
- a first (2) and a second (4) omnidirectional microphone located at a distance from each other which is smaller than a minimum acoustic wavelength defined by a given audio frequency range of operation,

- a first filtering means (6) for filtering a signal received by said first omnidirectional microphone and a second filtering means (8) for filtering a signal received by said second omnidirectional microphone, said first filtering means having a first frequency response yielding a first filtered output signal and said second filtering means having a second frequency response yielding a second filtered output signal, whereby said first and said second frequency response take into account the frequency responses of said first and second omnidirectional microphones,

- summing means (9) for summing said first and said second filtered output signal, so that the resulting summed signal has said first order directional pattern.
Microphone array system as in claim 1, wherein said first and second filtering means are implemented as finite-impulse response filters.
Microphone array system as in claim 2, comprising conversion means for converting the respective signals received by said first and second microphone into corresponding digital output signals,
Microphone array system as in claim 2 or 3, further comprising means for performing block processing for determining said signals output by said first and second filtering means.
Microphone array system as in any of the previous claims, wherein said first and said second filtering means are adaptable.
Microphone array system as in any of the previous claims, further comprising:
- a third omnidirectional microphone located at a distance from said first or said second microphone which is smaller than said minimum acoustic wavelength defined by said given audio frequency range of operation,

- a third filtering means for filtering a signal received by said third omnidirectional microphone, said third filtering means having a third frequency response yielding a third filtered output signal, whereby said third frequency response takes into account the frequency response of said third omnidirectional microphone,
and whereby said summing means is arranged for also adding said third filtered output signal.
Microphone array system as in any of claim 1 to 5, further comprising
- a third and a fourth omnidirectional microphone located at a distance from each other which is smaller than said minimum acoustic wavelength defined by said given audio frequency range of operation, said third and fourth omnidirectional microphone so positioned that they form an array with an axis substantially orthogonal with the axis of the array formed by said first and second omnidirectional microphone,

- a third filtering means for filtering a signal received by said third omnidirectional microphone and a fourth filtering means for filtering a signal received by said fourth omnidirectional microphone, said third filtering means having a third frequency response yielding a third filtered output signal and said fourth filtering means having a fourth frequency response yielding a fourth filtered output signal, whereby said third and said fourth frequency response take into account the frequency responses of said third and fourth omnidirectional microphones,
whereby said summing means is arranged for summing said third and fourth output signals and for combining the resulting output signal with said resulting output signal output by said first and second filtering means.
Microphone array system as in any of the previous claims, further comprising storage means for storing filter coefficient values for a plurality of steering angles.
Speaker unit of a conference system comprising a microphone array system as in any of the previous claims.
Speaker unit as in claim 9, arranged for being built-in in a table.
Speaker unit as in claim 10 arranged for being fold away.
Conference system comprising a plurality of speaker units as in any of claims 9 to 11.
Method for generating a microphone array output signal having a first order directional pattern, comprising the steps of
- providing a first and a second omnidirectional microphone located at a distance from each other which is smaller than a minimum acoustic wavelength defined by a given audio frequency range of operation,

- applying a signal received by said first microphone to a first filtering means and a signal received by said second microphone to a second filtering means, said first filtering means having a first frequency response and said second filtering means having a second frequency response, whereby said first and second frequency response take into account the frequency response of the first and second omnidirectional microphones,

- summing the signals output by said first and second filtering means, yielding said microphone array output signal with first order directional pattern.
Method for generating a microphone array output signal as in claim 13, wherein said first and second microphone are placed in end-fire configuration.