EP4661426A1

EP4661426A1 - Microphone system for capturing speech in an open-air vehicle

Info

Publication number: EP4661426A1
Application number: EP24179586.3A
Authority: EP
Inventors: Walter Daems; Jouke VERLINDEN; Jan STECKEL; Dennis LAURIJSSEN; Erik VERREYCKEN; Ralv LUNDAHL
Original assignee: Universiteit Antwerpen
Current assignee: Universiteit Antwerpen
Priority date: 2024-06-03
Filing date: 2024-06-03
Publication date: 2025-12-10
Also published as: WO2025252541A1

Abstract

Microphone system (10) mountable on an open-air vehicle (1) for capturing a speech signal (150) from a person, the system comprising a linear microphone array (100) comprising at least an outer left microphone (101) and an outer right microphone (104), the linear microphone array configured to be mounted on the open-air vehicle in front of the person; at least one out-of-line microphone (110) configured to be mounted between the linear microphone array and the person; and a control unit (170) configured to: obtain a first noise signal indicative for noise captured at the outer left microphone; obtain a second noise signal indicative for noise captured at the outer right microphone; and obtain a noisy speech signal; and subtract the first and second noise signals from the noisy speech signal thereby obtaining a filtered version of the speech signal.

Description

Technical field

The present invention generally relates to a microphone system mountable on an open-air vehicle, a computer-implemented method using the microphone system and an open-air vehicle comprising the microphone system.

Background

A hands-free device enables a user to have their hands available for other actions during use, i.e. it can be used without requiring a user to hold or otherwise operate the device.
In one application, hands-free devices are used during driving, e.g. in a car. This enables a driver to make calls or other voice recordings while keeping both hands on the steering wheel, which is required for safety reasons. As such, travel time can be used to perform business calls or other tasks involving speaking. Such a hands-free device may, for example, be a smartphone with an incorporated hands-free caller function and/or hands-free voice recording function.
The capturing of voice is performed by use of a microphone integrated into the hands-free device, placed at a distance from the driver. A problem with hands-free devices is that the microphone also captures noise originating from the surroundings. As long as the hands-free device is used in an enclosed space, such as a car, external noise may be fairly limited, but will become substantial under open-air circumstances. Therefore, hands-free devices that are tailored to function in an enclosed space, may not be suited for open-air applications, e.g. on a bicycle, a motorcycle or a convertible car. In the open air, a significant amount of wind noise can occur, especially during driving. In addition, the microphone is no longer shielded from other external noise, such as noise resulting from traffic.

Summary

The present invention aims to alleviate at least some of the above mentioned obstacles.
The scope of protection sought for various embodiments of the invention is set out by the independent claims.
The embodiments and features described in this specification that do not fall within the scope of the independent claims, if any, are to be interpreted as examples useful for understanding various embodiments of the invention.
According to some aspects, there is provided the subject matter of the independent claims. Some further aspects are defined in the dependent claims.
According to a first aspect, a microphone system for capturing a speech signal from a person is provided. The microphone system is mountable on an open-air vehicle. The microphone system comprises a linear microphone array, at least one out-of-line microphone and a control unit. The linear microphone array comprises at least an outer left microphone and an outer right microphone. The linear microphone array is configured to be mounted on the open-air vehicle in front of the person. The at least one out-of-line microphone is configured to be mounted between the linear microphone array and the person. The control unit is configured to: perform far-field beamforming using the linear microphone array in a first direction left to the travel direction to obtain a first noise signal indicative for noise captured at the outer left microphone; perform far-field beamforming using the linear microphone array in a second direction right to the travel direction to obtain a second noise signal indicative for noise captured at the outer right microphone; and perform near-field beamforming focused on the person by using the linear microphone array and the at least one out-of-line microphone to obtain a noisy speech signal; and subtract the first and second noise signals from the noisy speech signal thereby obtaining a filtered version of the speech signal.
The linear microphone array comprises at least two microphones. The outer left microphone is a first microphone configured to be mounted on a left side when looking along the travel direction. Analogously, the outer right microphone is a second microphone configured to be mounted on a right side when looking along the travel direction. The linear microphone array is a one-dimensional array of microphones, i.e. the at least two microphones may, for example, be arranged along a substantially straight line or a curved line.
The at least one out-of-line microphone is separate from the linear microphone array and is configured to be mounted outside of a line characterising a shape of the linear microphone array.
Beamforming includes delaying one or more microphone signals in time in accordance with a desired location-based restriction for focusing. Near-field beamforming includes determining such delays based on spherical sound wave propagation. Far-field beamforming includes determining delays based on planar sound wave propagation.
A near-field beamforming step comprises steering at least one out-of-line microphone and at least two of the microphones of the linear array towards a certain focal point, also referred to as steering point. The near-field beamforming can be focused on the person by setting a focal point corresponding to a position where a mouth of the person may be expected to be. Even though the near-field beamforming is focused on the person's mouth, surrounding noise signals from various directions are also captured. As a result, the noisy speech signal includes both speech as well as undesired surrounding noise.
The far-field beamforming steps comprise steering all the microphones of the linear microphone array towards the first and second directions, or corresponding steering angle, during post-processing of the captured microphone signals. By performing the far-field beamforming in the left direction, the noise captured on a left side of the microphone array is estimated. This noise component may mainly be captured by the outer left microphone. Analogously, by performing the far-field beamforming in the right direction, the noise captured on a right side of the microphone array is estimated. This noise component may mainly be captured by the outer right microphone. In an open-air vehicle, these noise estimates may contain a significant amount of wind noise. Other unwanted noise components in these directions are also captured, e.g. noise from surrounding traffic. By subtracting these noise estimates from the noisy speech signal, such noise components are removed from the noisy speech signal. On the other hand, sound coming from the direction of the person is not being cancelled out. As such, the captured speech signal is retained. This may improve the speech intelligibility such that the microphone system can be used in an open-air vehicle.
According to further example embodiments, the linear microphone array further comprises at least one inner microphone configured to be mounted between the outer left and outer right microphones.
The inner microphones may be located between the two outer microphones of the linear microphone array. By including more microphones in the linear microphone array, the far-field beamforming may become more accurate. When performing the far-field beamforming for both the first and second directions, all the microphones of the linear microphone array are used, i.e. both the outer and inner microphones. As more microphones are used, more statistical averaging out of errors in the received microphone signals can occur, e.g. due to microphone non-idealities. Also, more of a noise signal may be captured, e.g. due to the multipath acoustic environment resulting in some of the noise not reaching every microphone. One or more of the inner microphones can also be used for the near-field beamforming in conjunction with the out-of-line microphone. Therefore, more accurate estimates of the noise signals can be obtained, resulting in a less noisy speech signal.
According to further example embodiments, the control unit is further configured to, for one or more respective inner microphones: obtain an inner noise signal indicative for noise captured by the respective inner microphone by performing near-field beamforming focused on a position of the respective inner microphone using the linear microphone array; and further subtract the inner noise signal from the noisy speech signal.
Near-field beamforming allows a more accurate estimation of the noise captured by the inner microphones since the far-field assumption does not hold for the small distances between the microphones.
By subtracting inner noise signals, additional unwanted noise signals are cancelled. In particular, noise captured by the inner microphones of the linear array is removed. Therefore, a less noisy speech signal may be obtained.
According to further example embodiments, the control unit is further configured to obtain the inner noise signal by further using the at least one out-of-line microphone.
By including the at least one out-of-line microphone, an improved noise estimation of noise captured by the inner microphones may be obtained. The at least one out-of-line microphone may be able to capture a contribution of the noise captured by an inner microphone that the microphones of the linear microphone are not able to capture. The out-of-line microphone may capture such different contributions due to its out-of-line position where different signals are captured. In addition, by using more microphones for obtaining the inner noise signal, more statistical averaging out of errors may occur.
According to further example embodiments, the control unit is further configured to, for one or more respective out-of-line microphones: obtain an out-of-line noise signal indicative for noise captured by the respective out-of-line microphone by performing near-field beamforming focused on a position of the respective out-of-line microphone using the linear microphone array; and further subtract the out-of-line noise signal from the noisy speech signal.
By subtracting the one or more out-of-line noise signals, additional unwanted noise signals are cancelled. In particular, noise captured by the out-of-line microphones is removed. Therefore, a less noisy speech signal may be obtained.
According to further example embodiments, the control unit is further configured to: perform far-field beamforming using the linear microphone array in one or more additional directions different from the first and second directions to obtain one or more additional noise signals indicative for noise originating from the one or more additional directions respectively; and further subtract the one or more additional noise signals from the noisy speech signal.
By steering towards the additional direction(s), noise in directions other than the left and right directions can be cancelled as well. Thereby, additional wind noise components and/or other noise components, e.g. originating from surrounding traffic, coming from other directions can be removed.
According to further example embodiments, the control unit is further configured to adaptively determine one or more steering angles for the far-field beamforming, wherein a steering angle is an angle between an axis along the linear microphone array and a respective additional direction.
In other words, adaptive far-field beamforming may be implemented. Adaptive far-field beamforming may refer to updating the one or more steering angles over time. This allows tracking respective noise sources. This is especially useful since the open-air vehicle and/or noise sources can be moving. By performing adaptive far-field beamforming, the noise cancellation can thus be improved. Any one of the far-field beamforming steps can be made adaptive.
According to further example embodiments, the control unit is further configured to normalise at least one of the noise signals and the noisy speech signal before subtracting the noise signals from the noisy speech signal.
The normalising may include applying a gain coefficient to each of the at least one noise signals, wherein the gain coefficients sum up to one. By applying such a normalisation, the microphone system can compensate for unequal microphone characteristics of the microphones of the microphone system. In addition, the normalisation allows suppression of unwanted excessive noise compensation that takes away part of the useful speech signal.
According to further example embodiments, the far-field beamforming and the near-field beamforming comprises obtaining steered microphone signals.
According to further example embodiments, the control unit is further configured to perform the far-field beamforming and/or the near-field beamforming by further adaptively weighting the steered microphone signals so as to minimise noise in the filtered version of the noisy speech signal.
The adaptive weighting, i.e. applying adaptive filter coefficients, can be performed during any of the beamforming steps. By weighting the separate steered microphone signals, filtering can be performed along with the beamforming, e.g. filter-and-sum beamforming. Such filtering may be linear or non-linear. The filtering may, for example, be done by applying a finite impulse response, FIR, filter, an infinite impulse response, IIR, filter, or a machine learning mechanism, e.g. a neural network. When applying a linear filter, the differentiable characteristic thereof allows particularly quick computations. As such, the adaptive mechanism is able to react fast to changing noise circumstances. Additionally, a low latency can be achieved due to the limited calculation time required for the adaptive weighting.
According to further example embodiments, the control unit is further configured to jointly optimise the weighting and the normalising so as to minimise noise in the filtered version of the noisy speech signal.
Minimising noise in the filtered version of the noisy speech signal may be achieved by maximising the signal-to-noise ratio, SNR. By imposing a maximal SNR, the filtered version of the noisy speech signal will contain a maximal ratio of speech power to noise power. By combining the optimisation of the weighting and the normalising, a higher performance of the microphone system can be achieved.
According to further example embodiments, the control unit is further configured to adaptively determine a steering focal point for the near-field beamforming step that obtains the noisy speech signal, wherein the steering focal point is indicative for a position of the person.
Since the person can move around, the captured useful speech signal can be maximised by tracking the position of the person, e.g. the position of his or her mouth.
According to further example embodiments, the microphone system further comprises microphone housings configured to hold the at least one out-of-line microphone and the at least one outer left microphone and outer right microphone of the linear microphone array in respective recesses of the housings. A housing is shaped such that a separation point of airflow occurs in front of the respective microphone when the open-air vehicle is in motion.
Each microphone may be placed in a separate housing, or the microphones may be placed in a single housing providing the respective recesses.
A separation point, also referred to as breaking point, is the location where a flow of wind passing over an object releases from a surface thereof. At the separation point, the wind flow stops following a contour of the object. The position of the separation point may slightly fluctuate over time due to the dynamic nature of the wind flow. For example, such an object may be a steering wheel of a bicycle onto which the microphone system is attached. A wind flow may collide onto the steering wheel at the front of the steering wheel, i.e. the side facing the travel direction. After the separation point, a low pressure area emerges as a result from the wind disconnecting from the contour of the steering wheel. As a result, at the back of the steering wheel, i.e. on the driver's side, an area of turbulent wind flow can occur. The farther away from the separation point, the more turbulence may be created. A turbulent wind flow causes an increased effect of wind noise.
The housing is shaped such that the separation point occurs at a distance from the microphone. The microphone is placed near the separation point, yet avoiding the separation point itself and a marginal region around it. When placed at the separation point, the microphone may be subject to an alternating pressure as a result of the separation point fluctuations. Therefore, the housing is shaped such that it avoids the microphone being too close to the separation point. Further, the housing is shaped such that it avoids the microphone being too far from the separation point to avoid placing the microphone in a highly turbulent region.
By providing the separation point before the microphones, a low turbulence environment is created around the microphone inputs. As such, wind noise captured by the microphones is reduced.
According to further example embodiments, the microphones are covered by an open-cell foam layer, a hydrophobic layer, and a protective metal mesh layer.
The open-cell foam layer provides a damping layer against wind noise, while allowing speech signals to pass through. As such, the open-cell foam layer acts as a mechanical filter. Further, the hydrophobic layer prevents water from entering the microphones. The protective metal mesh layer protects the microphones from tear due to contact with external objects, e.g. scratches.
According to a second aspect, there is provided an open-air vehicle comprising the microphone system according to the first aspect. Such an open-air vehicle may provide one or more of the above-mentioned advantages.
An open-air vehicle is any vehicle wherein a traveller is at least partially exposed to the outside air surrounding the vehicle, e.g. a convertible car, a bicycle, an electric kick scooter, a motorcycle, a boat, a golf cart, or an agricultural vehicle without a windshield. In an open-air vehicle, the microphone system is susceptible to wind noise, in particular when the vehicle is in motion.
According to further example embodiments, the open-air vehicle is a bicycle and the microphone system is mounted on a handlebar of the bicycle.
Each microphone can be provided with a separate housing mounted on the steering wheel. Alternatively, the microphones can be integrated into the steering wheel.
According to a third aspect, a computer-implemented method is provided for capturing a speech signal from a person by using a microphone system mountable on an open-air vehicle, wherein the microphone system comprises: a linear microphone array comprising at least an outer left microphone and an outer right microphone, the linear microphone array configured to be mounted on the open-air vehicle in front of the person; at least one out-of-line microphone configured to be mounted between the linear microphone array and the person; and a control unit. The computer-implemented method comprises: performing far-field beamforming using the linear microphone array in a first direction left to the travel direction to obtain a first noise signal indicative for noise captured at the outer left microphone; performing far-field beamforming using the linear microphone array in a second direction right to the travel direction to obtain a second noise signal indicative for noise captured at the outer right microphone; performing near-field beamforming focused on the person by using the linear microphone array and the at least one out-of-line microphone to obtain a noisy speech signal; and subtracting the first and second noise signals from the noisy speech signal thereby obtaining a filtered version of the speech signal.
The microphone system may be a microphone system according to the first aspect. The steps of the computer-implemented method may be performed by the central unit of the microphone system.
According to a fourth aspect, a computer program is provided comprising instructions, which, when the program is executed by a computer, cause the computer to carry out the computer-implemented method according to the third aspect.
According to a fifth aspect, a computer-readable medium is provided comprising instructions which, when executed by a computer, cause the computer to carry out the computer-implemented method according to the third aspect.

Brief Description of the Drawings

Fig. 1 illustrates a microphone system mounted on a bicycle according to example embodiments;
Fig. 2 illustrates performing far-field beamforming using the linear microphone array according to example embodiments;
Fig. 3 illustrates performing near-field beamforming focused on a person to obtain a noisy speech signal according to example embodiments;
Fig. 4 illustrates open-air vehicles comprising the microphone system according to example embodiments;
Fig. 5 illustrates microphone housings for a microphone system according to example embodiments;
Fig. 6 illustrates a computer-implemented method for capturing a speech signal from a person by using a microphone system mountable on an open-air vehicle according to example embodiments;
Fig. 7 illustrates obtaining an inner noise signal indicative for noise captured by an inner microphone and subtracting the inner noise signal from the noisy speech signal according to example embodiments;
Fig. 8 illustrates performing far-field beamforming in an additional direction using the linear microphone array according to example embodiments; and
Fig. 9 illustrates a computing system for performing one or several steps of the computer-implemented method according to example embodiments.

Detailed Description of Embodiment(s)

The present disclosure relates to a microphone system operable in an open-air vehicle. Such a microphone system is configured to capture speech under the circumstances of wind noise and traffic noise in an open-air vehicle. For example, a voice signal uttered by a driver of the open-air vehicle can be captured. This allows the driver to make a phone call or to record voice without needing to let go of the steering wheel while speaking. Such a microphone system can be integrated into the vehicle during manufacturing or can be mounted thereon after manufacturing.
Figure 1 shows a microphone system 10 mounted on a bicycle 1 according to example embodiments. The bicycle 1 has: a saddle 145 for a driver to sit on; wheels 130, 131; pedals 141, 142 to actuate the wheels 130, 131; a steering wheel 165 for adjusting a driving direction 160; and a frame structure comprising various portions 120, 121, 122, 123, 124. The microphone system 10 is configured to capture a speech signal 150 from a driver sitting on the saddle 145 and is susceptible to noise caused by wind 180 surrounding the bicycle 1. During driving, the wind 180 may encounter the bicycle from the front, for example from the front towards the back, as indicated by arrows of the wind indications 180. The microphone system 10 is positioned in an area that is shadowed by the steering wheel 165, i.e. the headwind wind 180 does not impact the microphone system 10 directly. Tailwind would also not interfere with the operation of the microphone system 10, since the body of the driver of the bicycle 1 shields wind coming from behind. Sidewind may also only limitedly impact the microphone system 10 if the microphones are directed towards the driver. Appropriate shielding means may be further applied to this end.
The microphone system 10 comprises a linear microphone array 100 of four microphones 101, 102, 103, 104 mounted on the steering wheel 165 in front of a driver sitting on the saddle 145. The linear microphone array 100 is a one-dimensional array of microphones, i.e. the microphones 101-104 may, for example, be arranged along a substantially straight line or a curved line. The linear microphone array 100 is mounted on the steering wheel 165 or integrated therein. Therefore, the steering wheel determines the arrangement of the linear microphone array 100. The steering wheel 165 has a shape with slight curvatures. Alternatively, the steering wheel 165 and the linear microphone array 100 may have a more curved shape, or may be completely straight. The linear microphone array 100 has an outer left microphone 101 and an outer right microphone 104, wherein the terms 'left' and 'right' are defined with respect to the travel direction 160, i.e. as perceived by a driver sitting on the saddle 145. A left direction 191 and a right direction 192 are correspondingly defined.
The microphone system 10 further comprises at least one out-of-line microphone 110 placed outside an imaginary line connecting the linear microphone array 100. The out-of-line microphone 110 is mounted between the linear microphone array 100 and the saddle 145.
The microphone system 10 further comprises a control unit 170 configured to obtain a filtered version of the speech signal 150 by processing microphone signals captured by the microphones 101-104, 110. The control unit 170 is illustrated as a component separate from the microphones 101-104, 110, yet may also be incorporated into one or more of the microphones 101-104, 110. The control unit 170 may be connected, e.g. wirelessly, to an external device such as a smartphone. The control unit 170 may be configured to provide the filtered version of the speech signal to such an external device. As such, the control unit can make handsfree calling or use of other speech-related applications possible in the open-air vehicle 1.
The control unit 170 is configured to perform far-field beamforming using the linear microphone array 100 in the left direction 191 to obtain a first noise signal indicative for noise captured at the outer left microphone 101, e.g. including wind noise caused by the wind 180. The control unit 170 is further configured to perform far-field beamforming using the linear microphone array 100 in the right direction to obtain a second noise signal indicative for noise captured at the outer right microphone 104, e.g. including wind noise caused by the wind 180. The control unit 170 is further configured to perform near-field beamforming focused on a person sitting on the saddle 145 by using the linear microphone array 100 and the at least one out-of-line microphone 110 to obtain a noisy speech signal. The control unit 170 is further configured to subtract the first and second noise signals from the noisy speech signal thereby obtaining a filtered version of the speech signal 150. By doing so, undesired noise components including wind noise captured from the left and right sides of the bicycle 1 are removed. This may increase the speech intelligibility of the filtered speech signal.
Figure 2 shows a linear microphone array 200 having four microphones 201, 202, 203, 204 positioned on a straight line, e.g. a straight steering wheel 210 of a bicycle, according to example embodiments. A first noise source 21 originates from a first direction 291 to the left of the travel direction. A second noise source 22 originates from a second direction 292 to the right of the travel direction. An additional noise source 23 originates from an additional direction 293 different from the first and second directions 291, 292. Such a noise source may, for example, originate from surrounding traffic, e.g. an ambulance siren or roadworks. The right direction 292 can be defined as the direction at an angle of approximately 0° with respect to a line or axis defining the linear microphone array 200. The left direction 291 can correspondingly be defined as the direction at an angle α, 221, of approximately 180° with respect to the array axis. The left and right directions 291, 292 may be referred to as end-fire directions of the linear microphone array 200. A direction around an angle of 90°, e.g. where a driver of the open-air vehicle may be seated, may be referred to as a broadside direction of the linear microphone array 200. The additional direction 293 can, for example, be defined as the direction at an angle β, 223, of 120° with the array axis. It will be apparent that the additional noise source 23 can originate from an additional direction at any angle with respect to the array axis, apart from the already steered first and second directions, i.e. 0° and 180°. Optionally, even noise originating from the broadside direction where a driver is located can be obtained and subtracted from the noisy speech signal. To this end, voice activation detection, VAD, can be applied to distinguish between time periods when the microphone signals contain speech and when they don't. The noise originating from the broadside direction can then be obtained from speechless microphone signal segments.
A control unit of the microphone system to which the linear microphone array 200 pertains is configured to perform far-field beamforming using the linear microphone array 200 in the first direction 291 to obtain a first noise signal indicative for noise captured at the outer left microphone 201. By doing so, the first noise signal may originate from noise source 21. The control unit is further configured to perform far-field beamforming using the linear microphone array 200 in the second direction 292 to obtain a second noise signal indicative for noise captured at the outer right microphone 204. By doing so, the second noise signal may originate from noise source 22. The control unit may further be configured to perform far-field beamforming using the linear microphone array 200 in the additional direction 293 to obtain an additional noise signal indicative for noise originating from noise source 23. The noise signals are later subtracted from a noisy speech signal by the control unit.
Far-field beamforming includes making a far-field assumption, i.e. assuming that incoming sound waves are planar instead of spherical. Such an assumption is more accurate for sound sources farther away from the array. Planar sound waves arrive under the same angle of arrival at each of the microphones 201-204. During a far-field beamforming step, the control unit determines a direction to focus the linear microphone array 200 in the form of a steering angle. The steering angle is an angle between the axis along the linear microphone array and the chosen direction. For example, the control unit can steer the linear microphone array 200 towards one of the directions 291, 292, 293 with a steering angle of 0°, 180°, 120°, respectively. Due to the distances between the microphones 201-204, the signal will arrive at different moments in time at the microphones 201-204. The microphone signals are thus delayed with respect to each other. The delays depend on the distance between consecutive microphones 201 - 204 as well as the angle, a 221, β 223, at which a source 21 - 23 is located with respect to the array axis. To execute the far-field beamforming, the control unit realigns the incoming microphone signals of the microphones 201-204 such that these delays are compensated. To this end, the control unit may take into account the arrangement of the linear microphone array. For example, with a straight or substantially straight linear microphone array, the far-field beamforming may comprise delay-and-sum beamforming or filter-and-sum beamforming.
A microphone has fixed intrinsic directivity or directionality characteristics, i.e. from which directions the microphone can pick up signals and to what extent. The directivity can be expressed by a polar pattern illustrating an attenuation as a function of angle of direction. The directivity and corresponding polar pattern may, for example, be cardioid, hyper-cardioid, or omnidirectional. Such a polar pattern can also indicate the directivity of a microphone array as a whole. Because of the combined effect of the microphones 201-204, the array 200 can capture signals from all directions regardless of the polar patterns of the individual microphones. Thus, the polar pattern of the array 200 is at least zero for all angles and may be larger than zero for all angles. Figure 2 shows an ideal omnidirectional polar pattern 240 of the linear microphone array 200 having a gain of 1 for all angles, i.e. capturing signals from all directions without attenuating them. Optionally, the linear microphone array 200 may have a polar pattern that deviates from the ideal omnidirectional polar pattern 240, yet still having a gain larger than zero for all angles. The microphones 201-204 of the linear microphone array 200 can have a non-omnidirectional polar pattern, e.g. a cardioid polar pattern. The polar patterns of the individual microphones of the linear microphone array 200 may also differ from one another.
Figure 3 illustrates near-field beamforming to obtain a noisy speech signal according to example embodiments. Figure 3 further shows a three-dimensional view 311 of a bicycle 3 comprising the microphone system. The microphone system has a linear microphone array and an out-of-line microphone 310 according to example embodiments. The linear microphone array comprises microphones 301-305 and is mounted on a steering wheel 365 of the bicycle 3. The out-of-line microphone 310 is mounted on a frame of the bicycle 3 between the microphone 303 and a driver of the bicycle. A driver of the bicycle 3 is, under normal circumstances, seated on saddle 345. The mouth of the driver is, for example, located at position 350, which may be estimated during set-up of the microphone system.
During near-field beamforming, the spherical nature of sound waves is taken into account, contrary to far-field beamforming. Steering is performed towards a specific point in space instead of only a direction or steering angle. Near-field steering of a microphone signal of a single microphone includes considering sound sources located at a certain distance from the microphone. The steering is performed by applying a corresponding delay to the microphone signal, wherein the time delay and the distance are related by the velocity of sound. As such, during near-field beamforming of a single microphone, the microphone signal can be focused onto a sphere surface around the microphone. The sphere surface comprises all points in space that are at the same distance or radius from the microphone. The radius of the sphere surface is adjustable during the beamforming. For example, three-dimensional view 311 shows a sphere surface 361 around microphone 301. Microphone 301 is steered towards a distance equal to the radius of the sphere surface 361 during the near-field beamforming towards focal point 350. A person's mouth may, for example, be located at position 350. When only using microphone 301 to obtain the noisy speech signal, the microphone 301 also focuses on sound waves located elsewhere on the sphere surface 361. This may lead to obscuring of the speech signal.
Near-field beamforming of multiple microphones simultaneously allows focusing on a single point in space. For example, microphones 304 and 310 may also be used during the near-field beamforming. This results in focusing to respective sphere surfaces 362 and 363. By adjusting the radii of the sphere surfaces 361, 362, 363, the microphones 301, 304, 310 can be jointly steered towards a single focal point, i.e. to the intersection point 350 of the sphere surfaces. Two intersecting spheres define a circle. A third intersecting sphere may then define two points on such a circle: the intended focal point and a mirror point. Therefore, it takes at least three intersecting spheres to define distinct points in space. The mirror point of focal point 350 may be disregarded at least due to two reasons. Firstly, the mirror point occurs in a fairly silent region around the back of the front wheel of bicycle 3. Secondly, a directivity pattern and/or orientation of the microphones 301, 304, 310 may filter out sound at the mirror point. Thus, at least three microphones are required to obtain the noisy speech signal by focusing on a single point 350. Of those three microphones, at least one should preferably be an out-of-line microphone to provide a more extended range. The positions of microphones 301, 304, 310 restrict the possible focal points, since only intersection points of the corresponding spheres can be reached.
It will be apparent that the near-field beamforming may be performed using additional out-of-line microphones to further improve the spatial range. As such, more accurate speech signals may be obtained. For example, one or more additional out-of-line microphones may be mounted on a stem of the bicycle 3 or elsewhere on the frame of the bicycle 3.
Figure 3 further shows a top view 312 of the bicycle 3. The bicycle 3 comprises the microphone system having linear microphone array 300, an out-of-line microphone 310 and a control unit (not shown). The linear microphone array 300 comprises the microphones 301-305, i.e. outer left microphone 301, outer right microphone 305 and inner microphones 302, 303, 304. Microphone 303 is a centre microphone, i.e. located at a centre of the array 300. It is noted that centrally located microphone 303 is optional and it is not required to provide such a microphone 303. The speech source 350, the out-of-line microphone 310 and the centre microphone 303 are located in the same vertical plane shown in subplot 311. However, the driver of the bicycle 3 can make right and left movements during driving, thereby altering the position of speech source 350. The near-field beamforming may be performed adaptively to follow the speech source 350.
Figure 4 shows open-air vehicles comprising a microphone system according to example embodiments.
Top view 411 shows a roofless boat 45 according to example embodiments. The boat 45 travels in travel direction 460 and has a driver's seat 415. The boat 45 further comprises a microphone system 40. The microphone system 40 comprises a microphone array 400 of three microphones 401, 402, 403 mounted on a dashboard of the boat 45. The microphone system 40 further comprises an out-of-line microphone 410, e.g. mounted on a steering wheel or elsewhere on the dashboard of the boat.
Top view 412 shows a convertible car 46 according to example embodiments. The car 46 travels in travel direction 470 and has a driver's seat 435. The convertible 46 further comprises a microphone system 41. The microphone system 41 comprises a microphone array 420 of three microphones 421, 422, 423. The microphones 421-423 may be mounted on a dashboard 436 of the convertible 46. Alternatively, the microphones may be mounted on a steering wheel 437 of the convertible 46. In this case, a rotation angle of the steering wheel 437 may be measured during operation of the microphone system 41. The rotation angle is taken into account during capturing of speech in order to compensate for the effect of steering wheel movements on the beamforming. The microphone system 41 further comprises an out-of-line microphone 430, e.g. mounted on the front of the driver's seat 435.
Figure 5 illustrates a microphone housing for holding the microphones of the microphone system according to example embodiments. Three-dimensional view 501 shows a microphone housing 500 having a recess 560 for holding a single microphone of the microphone system. As such, each microphone can be placed in a separate housing 500. Alternatively, the microphones of the microphone system can be placed in a single housing providing the respective recesses. The housing 500 is mountable on an open-air vehicle, e.g. on a steering wheel of a bicycle. The housing has a U-shaped cross-section for clamping around an element of the open-air vehicle, e.g. around the steering wheel. The housing 500 can, for example, be mounted using fastening means, e.g. a clamp with a screw mechanism or zip-ties.
The housing 500 may be provided with a marking 580. Such a marking 580 can, for example, be used for automated identification of a position of the housing by the control unit during calibration of the microphone system. To this end, a user may take one or more pictures of the microphone system having a number of housings mounted on an open-air vehicle. From the one or more pictures, the control unit can identify the housing using the marking, by applying pattern recognition techniques. The control unit can further deduce the relative positioning of the housings and therefore of the microphones held within the housings. This information can be used to determine the exact arrangement of the microphone system and/or to aid the user to place the housings in such a way that the microphone system can perform optimally.
Front view 502 shows a housing 500 according to example embodiments. The housing 500 is made up of two symmetric portions 510, 520 that can be taken apart, e.g. for inserting or taking out a microphone. The recess 560 is correspondingly provided by two recesses 561, 562 of the portions 510, 520 respectively. The kink edge 550 is correspondingly provided by two edges 551, 552 of the portions 510, 520 respectively.
Three-dimensional view 503 shows a portion 520 according to example embodiments. Portion 510 may be provided as a symmetric version of portion 520. A cavity 590 is provided along a cross-section of the portion 520. Through this cavity 590, clamping means, e.g. a zip-tie or a string, can be provided for fixating the housing 500 onto the open-air vehicle, e.g. around a steering wheel. Optionally, the cavity 590 may be used for connecting electrical wires to the microphones and the control unit. Alternatively, the microphones and the control unit may communicate wirelessly.
The housing 500 is shaped such that a separation point of airflow occurs in front of the respective microphone when the open-air vehicle is in motion. To this end, a kink edge 550 is provided. In subfigure 504, the interaction of wind with a steering wheel 565 of a bicycle (not shown) is illustrated. Subfigure 504 shows a side-view of a steering wheel 565. Arrow 570 indicates the travel direction of the bicycle. From the perspective of the bicycle, the wind approaches in the opposite direction of the travel direction. The wind may have a laminar wind flow 541 before making contact with the steering wheel 565. The wind flow 541 may also be turbulent. Upon reaching the steering wheel 565, the wind flow 541 splits, indicated by point 545, to move along the contours of the steering wheel. At separation point 540, the wind flow 541 stops moving along the surface of the steering wheel 565. The deflection of the wind flow 541 at separation point 540 results in a turbulent region 542 occurring behind the steering wheel 565 as a result of a decreased pressure. In particular, at separation point 540, also referred to as breaking point 540, the wind flow 541 may cause the turbulent wind flow 542. The turbulence 542 may increase with distance from the separation point 540. Turbulent wind flow significantly increases captured wind noise and is therefore undesired to occur at an input of the microphones of the microphone system. It is noted that even with a laminar wind flow 541, a turbulent region 542 may occur behind the steering wheel.
Subfigure 505 illustrates the interaction of wind 541 with the steering wheel 565 having mounted thereon a housing 500 holding a microphone 506. Microphone 506 may, for example, be a microphone of a linear microphone array installed on the steering wheel 565. In subfigure 505, the same reference numerals are used to indicate the same functional components as indicated in three-dimensional view 501 and subfigure 504. As can be seen in subfigure 505, the kink edge 550 guides the wind flow 541 beyond the recess 560 holding the microphone 506. The separation point 540 therefore occurs at a distance from the microphone 506. As a result, the microphone 506 avoids the marginal region around fluctuating separation point 540 and exposure to noise created thereby. The turbulent region 542 has an increasing amount of turbulence farther away from the separation point 540. By the housing 500, the microphone 506 is located at an optimal position with regard to direct exposure to unwanted noise. A speech sound wave moving towards the microphone 506 may thus move through a low-turbulence region before reaching the microphone 506. As a result, the microphone 506 may capture a speech signal that contains less wind noise. This may lead to the noisy speech signal being more free of wind noise.
Figure 6 shows the steps a control unit is configured to perform according to example embodiments. A microphone system is illustrated comprising a linear microphone array consisting of microphones 601, 602, 603, 604. The microphone system further comprises an out-of-line microphone 610. The microphones 601, 602, 603, 604, 610 respectively obtain microphone signals 611, 612, 613, 614, 615. The control unit is not shown.
Referring to block 690, the control unit is configured to perform near-field beamforming 620 focused on a person. To this end, the control unit use one or more of the microphone signals 611-615 obtained by the linear microphone array 601-604 in combination with the out-of-line microphone 610 to obtain a noisy speech signal 637. In a first step 620, i.e. the near-field beamforming step, the microphone signals 611-615 are shifted in time to compensate for the corresponding delays, thereby obtaining steered microphone signals 621-625 respectively.
After the beamforming 620, the control unit may optionally apply weights 631, 632, 633, 634, 635 to the steered microphone signals 621, 622, 623, 624, 625 respectively, thereby obtaining weighted steered microphone signals 621', 622', 623', 624', 625'. The control unit sums the weighted or unweighted steered microphone signals 621-625, 621'-625', thereby obtaining a noisy speech signal 637. The weights 631-635, 651-654, 671-674 may also be referred to as filter coefficients.
Turning to block 691, the control unit is configured to perform far-field beamforming 640 in a first direction left to the travel direction. To this end, the control unit uses the microphone signals 611-614 obtained by the linear microphone array 601-604 to obtain a first noise signal 657. The first noise signal 657 is indicative for noise captured at the outer left microphone. In a first step 640, i.e. the far-field beamforming step, the microphone signals 611-614 are shifted in time to compensate for the delays in correspondence with a steering angle of 180°, thereby obtaining steered microphone signals 641-644 respectively.
After the beamforming 640, the control unit may optionally apply weights 651, 652, 653, 654 to the steered microphone signals 641, 642, 643, 644 respectively, thereby obtaining weighted steered microphone signals 641', 642', 643', 644'. The control unit sums the weighted or unweighted steered microphone signals 641-644, 641'-644', thereby obtaining the first noise signal 657.
In block 692, the control unit is configured to perform far-field beamforming 660 in a second direction right to the travel direction. To this end, the control unit uses the microphone signals 611-614 obtained by the linear microphone array 601-604 to obtain a second noise signal 677. The second noise signal 677 is indicative for noise captured at the outer right microphone. In a first step 660, i.e. the far-field beamforming step, the microphone signals 611-614 are shifted in time to compensate for the delays in correspondence with a steering angle of 0°, thereby obtaining steered microphone signals 661-664 respectively.
After the beamforming 660, the control unit may optionally apply weights 671, 672, 673, 674 to the steered microphone signals 661, 662, 663, 664 respectively, thereby obtaining weighted steered microphone signals 661', 662', 663', 664'. The control unit sums the weighted or unweighted steered microphone signals 661-664, 661'-664', thereby obtaining the first noise signal 677.
The noisy speech signal 637, the first noise signal 657 and the second noise signal 677 may be optionally weighted using weights 638, 658, 678 respectively to perform a normalisation. As a result, the control unit obtains a normalised noisy speech signal 639, a normalised first noise signal 659 and a normalised second noise signal 679 respectively.
The filter coefficients 631-635, 651-654, 671-674 may be implemented so as to include a desired frequency response during the steering. For example, a filter-and-sum beamformer 620, 640, 660 may be implemented using the coefficients 631-635, 651-654, 671-674. Another example may be a linearly constrained minimum variance, LCMV, beamformer 620, 640, 660, or a generalised sidelobe canceler, GSC.
In step 680, the control unit subtracts the noise signals 657, 677 or 659, 579 from the noisy speech signal 637 or 639 respectively, thereby obtaining a filtered version S, 681, of the speech signal.
The optional weighting steps using coefficients 631-635, 651-654, 671-674 may be iteratively adapted, i.e. regularly updated according to an optimisation function. In other words, the steered microphone signals may be weighted adaptively. The normalisation weights 638, 658, 678 may also be iteratively adapted according to an optimisation function. The coefficients and normalisation weights may optionally be jointly optimised according to a shared optimisation function. Further, the steering focal point of the near-field beamforming step 620 may also be iteratively adapted in order to track the speaker. The steering focal point may be jointly optimised along with the coefficients 631-635, 651-654, 671-674 and/or normalisation weights 638, 658, 678. A common optimisation function for these aspects may be expressed so as to maximise a signal-to-noise ratio, SNR, so as to minimise noise in the filtered version of the noisy speech signal.
Figure 7 shows the steps a control unit is configured to perform according to example embodiments. A microphone system is shown comprising a linear microphone array consisting of microphones 701, 702, 703, 704 and an out-of-line microphone 710. The microphones 701-704, 710 provide respective microphone signals 711-714, 715. In block 790, a noisy speech signal 770 is determined, e.g. by performing the steps of block 690 of Figure 6. In block 791, the first noise signal 771 is determined, e.g. by performing the steps of block 691 of Figure 6. In block 792, the second noise signal 772 is determined, e.g. by performing the steps of block 692 of Figure 6. Figure 7 includes the steps as illustrated in Figure 6 and illustrates how noise captured at the inner microphones 702, 703 can additionally be removed from the noisy speech signal.
Figure 7 shows block 793, wherein, for one of the inner microphones 702, 703 of the linear microphone array, an inner noise signal 750 is obtained. The inner noise signal is indicative for noise captured by the respective inner microphone. In block 793, the control unit performs near-field beamforming 720 focused on a position of the respective inner microphone using the linear microphone array. Optionally, the control unit may also use the microphone signal captured by the out-of-line microphone.
After the beamforming 720, the control unit may optionally apply weights 731, 732, 733, 734, 735 to the steered microphone signals 721, 722, 723, 724, 725 respectively, thereby obtaining weighted steered microphone signals 721', 722', 723', 724', 725'. The control unit then sums the weighted or unweighted steered microphone signals 721-725, 721'-725', thereby obtaining the inner noise signal 750. The weights 731-735 may also be referred to as filter coefficients. Optionally, the inner noise signal may be included in the normalisation performed on the blocks 790, 791, 792 by applying a weight 760. Thereby, the control unit obtains a normalised inner noise signal 773.
In step 780, the control unit subtracts the noise signals 771, 772 and 750 or 773 from the noisy speech signal 770, thereby obtaining a filtered version S, 781, of the speech signal.
It will be apparent that the steps of block 793 may be performed for each of the inner microphones 702, 703. Further, the steps of block 793 may also be performed for one or more of the out-of-line microphones, i.e. microphone 710. In that case, an out-of-line noise signal 750 is obtained indicative for noise captured by the out-of-line microphone 710. The near-field beamforming 720 is then focused on a position of the out-of-line microphone 710 using the linear microphone array 701-704.
Figure 8 shows the steps a control unit is configured to perform according to example embodiments. A microphone system is shown comprising a linear microphone array consisting of microphones 801, 802, 803, 804 and an out-of-line microphone 810. The microphones 801-804, 810 provide respective microphone signals 811-814, 815. In block 890, a noisy speech signal 870 is determined, e.g. by performing the steps of block 690 of Figure 6. In block 891, the first noise signal 871 is determined, e.g. by performing the steps of block 691 of Figure 6. In block 892, the second noise signal 872 is determined, e.g. by performing the steps of block 692 of Figure 6. In block 893, an inner noise signal 873 is determined for an inner microphone 802, 803. This block may be repeated for the remaining inner microphone. As such, Figure 8 includes the steps as illustrated in Figures 6 and 7 and illustrates how noise originating from an additional direction can additionally be removed from the noisy speech signal.
Further, Figure 8 shows a block 894 for obtaining an additional noise signal 850 indicative for noise originating from a direction additional to the left and right directions. In block 894, the microphone signals 811, 812, 813, 814 captured by the microphones 801, 802, 803, 804 of the linear microphone array are steered towards the additional direction, e.g. with steering angle 120°. As a result, steered microphone signals 821-824 are obtained.
After the beamforming 820, the control unit may optionally apply weights 831, 832, 833, 834 to the steered microphone signals 821, 822, 823, 824 respectively, thereby obtaining weighted steered microphone signals 821', 822', 823', 824'. The control unit then sums the weighted or unweighted steered microphone signals 821-824, 821'-824', thereby obtaining the inner noise signal 850. The weights 831-834 may also be referred to as filter coefficients. Optionally, the inner noise signal may be included in the normalisation performed on the blocks 890, 891, 892, 893 by applying a weight 860. Thereby, the control unit obtains a normalised additional noise signal 874.
It is noted that during any of the far-field beamforming steps, the microphone signal obtained by an out-of-line microphone 810 may optionally also be taken into account when combined with adaptive weighting using weights 831-835. The microphone signal 815 is then also steered towards the additional direction, resulting in steered microphone signal 825. The signal 825 is then weighted with the adaptive weight or filter coefficient 835, resulting in the weighted steered microphone signal 825'. By doing so, the out-of-line microphone can provide a useful contribution to the noise signal due to the imposed optimisation criterion. Additionally, the steering angle of the out-of-line microphone 810 may be implemented as adaptive and initialised with the steering angle that is used for the linear microphone array. By initialising the out-of-line microphone 810 in this manner, the out-of-line microphone 810 may evolve towards steering to the same direction as the array due to the adaptive mechanism.
Further, the steering angle of the far-field beamforming step 820 may be iteratively adapted in order to adapt to changing or moving noise sources. The steering angle may be jointly optimised along with the filter coefficients and/or normalisation weights. This may be included in a common optimisation function that imposes maximisation of the signal-to-noise ratio, SNR, so as to minimise noise in the filtered version of the noisy speech signal.
Figure 9 shows a suitable computing system 900 enabling to implement embodiments of the microphone system, e.g. the control unit. Computing system 900 may in general be formed as a suitable general-purpose computer and comprise a bus 910, a processor 902, a local memory 904, one or more optional input interfaces 914, one or more optional output interfaces 916, a communication interface 912, a storage element interface 906, and one or more storage elements 908. Bus 910 may comprise one or more conductors that permit communication among the components of the computing system 900. For example, the conductors may comprise one or more wireless conductors. Processor 902 may include any type of conventional processor or microprocessor that interprets and executes programming instructions. Local memory 904 may include a random-access memory (RAM) or another type of dynamic storage device that stores information and instructions for execution by processor 902 and/or a read only memory (ROM) or another type of static storage device that stores static information and instructions for use by processor 902. Input interface 914 may comprise one or more conventional mechanisms that permit an operator or user to input information to the computing device 900, such as a keyboard 920, a mouse 930, a pen, voice recognition and/or biometric mechanisms, a camera, etc. Output interface 916 may comprise one or more conventional mechanisms that output information to the operator or user, such as a display 940, etc. Communication interface 912 may comprise any transceiver-like mechanism such as for example one or more Ethernet interfaces that enables computing system 900 to communicate with other devices and/or systems, for example with the microphones 101-104, 110 of the microphone system and a client device of the person such as a smartphone. The communication interface 912 may, for example, be a Bluetooth Classic or a Bluetooth Low Energy communication interface. The communication interface 912 of computing system 900 may be connected to such another computing system by means of a local area network (LAN) or a wide area network (WAN) such as for example the internet. Storage element interface 906 may comprise a storage interface such as for example a Serial Advanced Technology Attachment (SATA) interface or a Small Computer System Interface (SCSI) for connecting bus 910 to one or more storage elements 908, such as one or more local disks, for example SATA disk drives, and control the reading and writing of data to and/or from these storage elements 908. Although the storage element(s) 908 above is/are described as a local disk, in general any other suitable computer-readable media such as a removable magnetic disk, optical storage media such as a CD or DVD, -ROM disk, solid state drives, flash memory cards, ... could be used. Computing system 900 could thus be configured to comprise program instructions stored thereon for performing the computer-implemented method according to the first example aspect.
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

Microphone system (10) mountable on an open-air vehicle (1) for capturing a speech signal (150) from a person, the system comprising:
- a linear microphone array (100) comprising at least an outer left microphone (101) and an outer right microphone (104), the linear microphone array configured to be mounted on the open-air vehicle in front of the person;

- at least one out-of-line microphone (110) configured to be mounted between the linear microphone array and the person; and

- a control unit (170) configured to:
- perform far-field beamforming (640) using the linear microphone array in a first direction (191) left to the travel direction to obtain a first noise signal (657) indicative for noise captured at the outer left microphone;

- perform far-field beamforming (660) using the linear microphone array in a second direction (192) right to the travel direction to obtain a second noise signal (677) indicative for noise captured at the outer right microphone; and

- perform near-field beamforming (620) focused on the person by using the linear microphone array and the at least one out-of-line microphone to obtain a noisy speech signal (637); and

- subtract (680) the first and second noise signals from the noisy speech signal thereby obtaining a filtered version (681) of the speech signal.
The system according to claim 1, wherein the linear microphone array further comprises at least one inner microphone (102, 103) configured to be mounted between the outer left and outer right microphones.
The system according to claim 2, wherein the control unit is further configured to, for one or more respective inner microphones:
- obtain an inner noise signal (750) indicative for noise captured by the respective inner microphone by performing near-field beamforming (720) focused on a position of the respective inner microphone using the linear microphone array; and

- further subtract (780) the inner noise signal from the noisy speech signal.
The system according to claim 3, wherein the control unit is further configured to obtain the inner noise signal by further using the at least one out-of-line microphone.
The system according to any one of the preceding claims, wherein the control unit is further configured to, for one or more respective out-of-line microphones:
- obtain an out-of-line noise signal (750) indicative for noise captured by the respective out-of-line microphone by performing near-field beamforming (720) focused on a position of the respective out-of-line microphone using the linear microphone array; and

- further subtract (780) the out-of-line noise signal from the noisy speech signal.
The system according to any one of the preceding claims, wherein the control unit is further configured to:
- perform far-field beamforming (820) using the linear microphone array in one or more additional directions (493) different from the first and second directions to obtain one or more additional noise signals (850) indicative for noise originating from the one or more additional directions respectively; and

- further subtract (880) the one or more additional noise signals from the noisy speech signal.
The system according to claim 6, wherein the control unit is further configured to adaptively determine one or more steering angles for the far-field beamforming (820), wherein a steering angle is an angle between an axis along the linear microphone array and a respective additional direction.
The system according to any one of the preceding claims, wherein the control unit is further configured to normalise (638, 658, 678) at least one of the noise signals and the noisy speech signal before subtracting the noise signals from the noisy speech signal.
The system according to any of the preceding claims, wherein the far-field beamforming (640, 660, 820) and the near-field beamforming (620, 720) comprises obtaining steered microphone signals (641-644, 661-664, 821-825, 621-625, 721-725), and wherein the control unit is further configured to perform the far-field beamforming and/or the near-field beamforming by further adaptively weighting the steered microphone signals so as to minimise noise in the filtered version of the noisy speech signal.
The system according to claim 8 or 9, wherein the control unit is further configured to jointly optimise the weighting and the normalising so as to minimise noise in the filtered version of the noisy speech signal.
The system according to any one of the preceding claims, wherein the control unit is further configured to adaptively determine a steering focal point for the near-field beamforming step that obtains the noisy speech signal, wherein the steering focal point is indicative for a position of the person.
Open-air vehicle comprising the microphone system according to any one of the preceding claims.
Open-air vehicle according to claim 13, wherein the open-air vehicle is a bicycle and wherein the microphone system is mounted on a handlebar of the bicycle.
Computer-implemented method for capturing a speech signal (150) from a person by using a microphone system mountable on an open-air vehicle, the microphone system comprising:
- a linear microphone array (100) comprising at least an outer left microphone (101) and an outer right microphone (104), the linear microphone array configured to be mounted on the open-air vehicle in front of the person;

- at least one out-of-line microphone (110) configured to be mounted between the linear microphone array and the person; and

- a control unit (170);
wherein the computer-implemented method comprises:
- performing far-field beamforming (640) using the linear microphone array in a first direction (191) left to the travel direction to obtain a first noise signal (657) indicative for noise captured at the outer left microphone;

- performing far-field beamforming (660) using the linear microphone array in a second direction (192) right to the travel direction to obtain a second noise signal (677) indicative for noise captured at the outer right microphone;

- performing near-field beamforming (620) focused on the person by using the linear microphone array and the at least one out-of-line microphone to obtain a noisy speech signal (637); and

- subtracting (680) the first and second noise signals from the noisy speech signal thereby obtaining a filtered version (681) of the speech signal.
The microphone system according to any one of the claims 1-11, further comprising microphone housings configured to hold the at least one out-of-line microphone and the at least one outer left microphone and outer right microphone of the linear microphone array in respective recesses of the housings; wherein the housing is shaped such that a separation point (540) of airflow occurs in front of the respective microphone when the open-air vehicle is in motion.