Field of Invention

The present invention relates to noise reduction in speech communication and, in particular, to the reduction of wind noise in hands-free sets.

Prior Art

Microphones are subject to environmental noise and, in particular, non-acoustic perturbations as wind noise. Wind noise is, e.g., caused by vibrations of the diaphragm of the microphone due to wind. By air turbulence in the vicinity of the microphone low frequency noise, as in the form of popps, is caused that deteriorates the microphone signals. Consequently, speech signals detected by the microphone and sent to a receiver can become unintelligible.

Wind and vibrational noise represent a cumbersome problem in outdoor use of microphones and in car communication systems used for telephony or speech recognizing means, in particular, if hands-free sets are employed. Vibrational noise in microphones installed in vehicular cabins result, e.g., from engine vibration.

In vehicles, directional microphones are used to reduce noise in dependence of the direction of noise sources. On the other hand, directional microphones are rather sensitive to wind noise. Wind noise can be caused by blowers. In some vehicles microphones are placed in a way that the air ejected from blower jets may hit the microphones directly. Open windows, an open sliding roof and an open cabriolet hood represent other sources of wind noise.

One method known in the art to reduce wind noise is to cover microphones by layers of foam. Other mechanical kinds of wind protection means are available. In order to be sufficiently effective, the mechanical wind protection means, however, have to be made relatively bulky. This is particularly disadvantageous in the context of microphones installed in vehicular cabins, since the cabin design and available space simply does not allow for bulky wind protection means around microphones. Moreover, windbreaks tend to degrade the overall performance of the microphones, in particular, of directional microphones. The directivity of the latter usually is reduced when mechanical wind protecting means are applied.

Thus, it is an object of the present invention to provide a microphone system with an improved reduction of non-acoustic perturbations and a method for reduction of non- acoustic perturbations in microphone signals and, in particular, to provide a microphone system with an improved wind noise reduction that can be installed in a vehicular cabin as part of a handsfree set.

Description of the invention

The above mentioned object is achieved by a microphone system, comprising

at least one directional microphone with a first directivity to obtain a first microphone signal;

at least one directional microphone with a second directivity that is worse (lower) than the first directivity, in particular, an omni-directional microphone, to obtain a second microphone signal;

a detection means configured to receive the first and the second microphone signals and to output a detection signal indicating whether non-acoustic perturbations, in particular, wind noise, in the first and/or the second microphone signals exceed a predetermined level; and

a signal processing means configured to output an output signal generated from the first microphone signal and/or the second microphone signal on the basis of the detection signal.

At least two microphones with different directivities are use. The at least one microphone with the second directivity may be an omni-directional microphone with one single inlet port for sound to enter. Such a microphone senses the instantaneous acoustic pressure at the sound port regardless of where the sound sources are located. An ideal omni-directional microphone will exhibit a polar sensitivity pattern, which is a perfect circle. An omni-directional microphone can be considered as being relatively wind robust, i.e. insensitive to wind noise.

The at least one microphone with the first directivity one shows a directivity index that significantly differs from zero. The directivity index is defined by the log ratio of the power delivered by an omni-directional microphone to that of a directional microphone with equal sensitivity in the principle direction, in a diffuse sound field.

The at least one microphone with the first directivity may be a directional microphone with, e.g., a cardioid or a dipole polar pattern and a directivity index of approximately 4.8 dB.

Non-acoustic perturbations comprise wind noise, vibrational noise, resonances, etc. In particular, the omni-directional microphone may be optimized for wind robustness and may also be provided with additional mechanical wind protection means, e.g., a foam layer. The detection means may physically and/or logically be part of the signal processing means.

Detection of non-acoustic perturbations, in particular, wind noise, above or below a predetermined level may be performed on the basis of the (short-time) power of microphone signals comprising non-acoustic perturbations. For example, the power of the signal from the microphone with the first directivity is compared with the one of the signal of the microphone with the second directivity. If the signal from the (wind robust) microphone with the second (worse) directivity shows a lower power level by

some predetermined value than the one from the microphone with the first (better, higher) directivity, it can be determined that non-acoustic perturbations, in particular, wind noise are present.

Thus, the detection means may comprise a comparator configured to compare the power level of the first microphone signal with the one of the second microphone signal and the detection signal may be generated and output on the basis of the comparison result.

The predetermined level above which wind noise significantly deteriorates the microphone signals can be determined empirically and/or on the basis of theoretical calculations. Depending on the specification of the application this level can be chosen to be close to zero. It may also be preferred to provide a selection means, by which a user may select from several levels thereby controlling the sensitivity of the microphone system with respect to the suppression of wind noise.

The signal processing means may at first A/D convert the microphone signals before any further processing. In general, the processing may be performed in the time domain, the frequency domain (after a Fourier transformation) or the subband domain. The signal processing may also be done with analogue signals without using an A/D or D/A converter, respectively.

According to the invention, it is determined whether or not non-acoustic perturbations are present at some given level. The output signal is generated according to the result of this determination process. Since a signal obtained by a microphone with a better directivity is usually more deteriorated by, e.g., wind noise, than a signal obtained by a microphone with a worse directivity, if wind noise is detected, the generation of the output signal may be primarily be carried out taken into account the second microphone signal.

The herein disclosed microphone system shows an improved reduction of non- acoustic perturbations as compared to the prior art. Reliable operation is guaranteed by the generation of an output signal from microphone signals obtained by micro-

phones of different directivity in dependence on the level of non-acoustic perturbations. The microphone system is particularly suitable for installation in a vehicular cabin. In this case, the number of microphones used may be rather limited.

According to a relatively simple but robust embodiment, the output signal can exclusively be generated from the first microphone signal, i.e. without any contributions from the second microphone signal, if the detection signal indicates that the non- acoustic perturbations do not exceed the predetermined level. The higher quality of the microphone(s) with a higher directivity can be fully used in the case of non- disturbing wind noise, for example.

On the other hand, the output signal may be exclusively generated from the second microphone signal, i.e. without any contributions from the first microphone signal, if the detection signal indicates that the non-acoustic perturbations exceed the predetermined level.

According to a further embodiment of the inventive microphone system the output signal can be generated from a portion of the first microphone signal in a frequency range above a predetermined frequency and from a portion of the second microphone signal in a frequency range below the predetermined frequency, if the detection signal indicates that the non-acoustic perturbations exceed the predetermined level or if no detection signal is output. Even when no detection signal is output at all, the microphone signals may be combined in order to achieve a higher quality. The predetermined frequency may be chosen as 500 Hz, for example.

A more elaborated embodiment of the inventive microphone system further comprises a frequency determining means configured to determine time-dependently a threshold frequency above that only non-acoustic perturbations that do not exceed the predetermined level are detected and wherein the output signal is generated from a portion of the first microphone signal in a frequency range above the threshold frequency and from a portion of the second microphone signal in a frequency range below the threshold frequency.

Thus, instead of simply switching between the microphone(s) with the first directivity and the microphone(s) with the second directivity to be used for generating the output signal, both kinds of microphones may be used. The frequency determining means may physically and/or logically be part of the signal processing means.

Since wind noise, e.g., mainly affects the low frequency range, the high frequency range of the output signal above a given threshold frequency may be generated from the first microphone signal regardless of the level of wind noise. On the other hand, the portion of the output signal with a frequency range below the threshold frequency, i.e. in the low frequency range, may be generated from the second microphone signal obtained by the wind robust microphone. The width of the perturbed frequency range, however, depends on the level of the wind noise. In general, depending on the level of wind perturbations, the threshold frequency may be some 500 Hz to some 1000 Hz.

In some detail, the frequency dependent processing can, e.g., be carried out as follows. The microphone signals are (short-time) Fourier transformed or alternatively processed in subbands (with some undersampling). The resulting complex spectrograms can be further processed with respect to both amplitude and phase. Below the threshold frequency only the phase values of the second microphone signal may be used. As for the amplitude of the output signal for each spectral value the minimum of the respective spectral value of the microphone signal from the at least one microphone with the first directivity and the spectral value of the microphone from the at least one microphone with the second directivity may be used. This allows for automatically selecting for each frequency the amplitude with the lower perturbation for the generation of the output signal. For low frequency portions that do not show wind noise the signal amplitude of the at least one microphone with the first directivity can be used, since this microphone exhibits a better general reduction of background noise than the one with the second directivity.

Thus, according to an embodiment of the inventive microphone system the signal processing means is configured to generate the output signal for frequencies below the threshold frequency on the basis of the phase values of the second microphone signal and/or amplitude values representing for each frequency the minimum of the

respective spectral value of the first microphone signal and the respective spectral value of the second microphone signal. In particular, the signal processing means may be configured to generate an output signal that exhibits the same phase values as the second microphone signal and the same amplitudes per frequency as the minimum of the respective spectral value of the first microphone signal and the respective spectral value of the second microphone signal for frequencies below the threshold frequency.

Since the inventive microphone system comprises different kinds of microphones, in general, the frequency responses of the microphones have to be adjusted to each other. The adjustment may be carried out adaptively and preferably when no non- acoustic perturbations are present.

In order to avoid artifacts in the output signal caused by abrupt changes in the threshold frequency and/or the determination whether or not significant non-acoustic perturbations are present, a detection delay means configured to keep the detection signal constant over a predetermined period and also a threshold frequency delay means configured to keep the threshold frequency constant over a predetermined period may be incorporated in embodiments of the microphone system.

Depending on the desired quality of the output signal more than one directional microphone with directivities better than the second directivity and a beamforming means configured to receive microphone signals from each of the directional microphones with a directivity better than the second directivity and to output a beamformed microphone signal that represents the first microphone signal generated from the microphone signals from each of the directional microphones with a directivity better than the second directivity may be employed.

The beamforming means combines the microphone signals from each of the directional microphones with a directivity better than the second directivity to obtain a beamformed signal with an increased signal-to-noise ratio as known in the art. If the microphones are used, e.g., to detect voice, the speech signals are deteriorated by acoustic background noise. Speech signals and noise signals overlap in the time and

frequency domain. By using multiple microphones the different spatial characteristics of speech and noise can be exploited and in this way the background noise can be suppressed. In the disclosed microphone system, e.g., a delay-and-sum beamformer or a more elaborated beamformer combined with adaptive filtering (Griffiths-Jim beamforming) as known in the art may be used.

Furthermore, the microphone system according to one of the above described embodiments may comprise a single housing containing each of the microphones.

The invention also provides a handsfree set comprising the microphone system according to one of the above-described embodiments. In particular, in hands-free sets installed in vehicular cabins the intelligibility of speech signals can be improved by employment of the embodiments of the inventive microphone system.

The above mentioned object is also achieved by a method for reducing non-acoustic perturbations in microphone signals, comprising

detecting an audio signal by at least one directional microphone of a first directivity to obtain a first microphone signal;

detecting the audio signal by at least one directional microphone with a second directivity that is worse (lower) than the first directivity, in particular, an omni-directional microphone, to obtain a second microphone signal;

determining whether non-acoustic perturbations, in particular, wind noise, that exceed a predetermined level are present in the first and/or the second microphone signal; and

outputting an output signal generated from the first microphone signal and/or the second microphone signal on the basis of the result of the determining whether non- acoustic perturbations that exceed a predetermined level are present in the first and/or the second microphone signal.

The output signal may exclusively be generated from the first microphone signal, i.e. without any contribution from the second microphone signal, if non-acoustic perturbations that do exceed the predetermined level are not present and/or the output signal may exclusively be generated from the second microphone signal, i.e. without any contribution from the first microphone signal, if non-acoustic perturbations that exceed the predetermined level are present.

According to another example of the inventive method the output signal can be generated from a portion of the first microphone signal in a frequency range above a predetermined frequency and from a portion of the second microphone signal in a frequency range below the predetermined frequency, if the detection signal indicates that the non-acoustic perturbations exceed the predetermined level or if no detection signal is output.

Furthermore, a threshold frequency may be determined time-dependently above that only non-acoustic perturbations that do not exceed the predetermined level are detected and the output signal may be generated from a portion of the first microphone signal in a frequency range above the threshold frequency and from a portion of the second microphone signal in a frequency range below the threshold frequency.

The time- and frequency dependent processing results in a particularly efficient reduction of non-acoustic perturbations.

After a Fourier transformation or a subband decomposition of the microphone signals the phase values of the second microphone signal and/or amplitude values representing for each frequency the minimum of the respective spectral value of the first microphone signal and the respective spectral value of the second microphone signal may be used for the generation of the output signal for frequencies below the threshold frequency. In particular, the output signal may exhibit the same phase values as the second microphone signal and the same amplitudes per frequency as the minimum of the respective spectral value of the first microphone signal and the respective spectral value of the second microphone signal for frequencies below the threshold frequency.

It may be preferred that more than one directional microphone with a directivity better than the second directivity detect the audio signals and that the microphone signals from each of the directional microphones with directivities better than the second directivity are beamformed to output a beamformed microphone signal as the first microphone signal, in order to further enhance the quality of the output signal.

The result of the step of determining whether non-acoustic perturbations that exceed a predetermined level are present in the first and/or the second microphone signal may be kept constant over a predetermined period and/or the threshold frequency may be kept constant over another predetermined period that is usually different for the first mentioned period.

In order to avoid abrupt changes in the generation of output signals, a change from the first microphone signal to the second microphone signal as a basis for the generation of the output signal, e.g., may not be performed immediately, if wind noise is detected, whereas it was not detected before or vice versa. In particular, if suddenly wind noise is detected that was not present before, for some period, e.g., for some seconds, the processing is performed, as if still no wind noise (or wind noise below a predetermined level only) was present. Similar, if the determined threshold frequency suddenly changes, the combination from portions of the first and second microphone signals to gain the output signal is modified only after some period.

In the above mentioned embodiments of the herein disclosed method determining whether non-acoustic perturbations, in particular, wind noise, that exceed a predetermined level are present in the first and/or the second microphone signal may comprise comparing the power level of the first microphone signal with the power level of the second microphone signal (for details see above).

The invention also provides a computer program product comprising one or more computer readable media having computer-executable instructions for performing the steps of embodiments of the method for reducing non-acoustic perturbations in microphone signals as described above.

Additional features and advantages of the invention will be described with reference to the drawings. Similar elements in different Figures have been given the same reference numeral for purposes of consistency and simplicity.

Figure 1 shows an example of the inventive microphone system comprising one directional and one omni-directional microphone.

Figure 2 shows an example of the inventive microphone system comprising three directional microphones, one omni-directional microphone and a beam- former.

Figure 3 shows steps of an example of the inventive method for reducing wind noise in microphone signals comprising comparing the signal power of the microphone signals and outputting a signal of a directional microphone, if no wind noise is detected.

Figure 4 shows steps of an example of the inventive method for reducing wind noise in microphone signals comprising the determination of a threshold frequency and outputting a signal combined from the microphone signal from a directional microphone and the microphone signal from an omni-directional microphone.

A relatively simple version of the herein disclosed microphone system is illustrated in Figure 1. A directional microphone 1 and an omni-directional microphone 2 are contained in one single microphone housing 3. The omnidirectional microphone 2 can be provided with some mechanical wind protection means as a foam layer (in Figure 1 this is indicated by the shading). Both microphones detect audio signals and the corresponding microphone signals are input in a signal processing means 4. The signal processing means 4 comprises a detection means for determining whether or not wind noise above a predetermined level is present.

The check for significant wind noise is performed by comparing the short-time power of the microphone signals from the directional microphone 1 and the omni-directional microphone 2 with each other after having the frequency responses of the microphones adjusted to each other. A lower power of the microphone signal obtained by the omni-directional microphone 2 indicates the presence of wind noise. The level of wind noise can be estimated from the difference in power of the both microphone signals (the one from the directional microphone 1 and the one from the omni-directional microphone 2).

If no wind noise or wind noise below a predetermined level only is detected in the microphone signals the signal processing means 4 outputs the microphone signals obtained by the directional microphone. If, on the other hand, wind noise above the predetermined level is detected, microphone signals obtained by the omni-directional microphone are output by the signal processing means 4.

The switching between output of microphone signals obtained by the directional microphone to output of microphone signals obtained by the omnidirectional microphone in dependence of wind noise is not carried out abruptly, if the wind noise exceeds or falls below the predetermined level. Rather, a counter is employed that starts counting once the wind noise exceeds the predetermined level and only after a predetermined number of counts detection of wind noise only below the predetermined level results in an output of microphone signals obtained by the directional microphone.

According to this example, the output signals are transmitted to a speech recognition means 5. The recognizing results obtained by the speech recognition means 5 are more reliable than in the case in that conventional microphone systems suffering from wind noise are used. The illustrated example of a microphone system comprising microphones 1 and 2, a microphone housing 3 and a signal processing means 4 is advantageously installed in a vehicular cabin to facilitate speech conversation.

Figure 2 shows a more elaborated example of the inventive microphone system. Here, a number of directional microphones 1 , e.g. three directional microphones, and one omni-directional microphone 2 are employed. The housing 3 may also contain more than one omni-directional microphone 2. The omnidirectional microphone 2 shows a less wind noise sensitivity as compared to the directional microphones 1. The latter may have different directivites.

The microphone signals of the directional microphones are input in a beam- former 6 that may be either a fixed beamformer or an adaptive one as they are known in the art.

The beamformer 6 outputs a beamformed microphone signal with an increased signal-to-noise ratio. The beamformed microphone signal as well as the microphone signal obtained by the omni-directional microphone are input in the signal processing means 4 that operates in a way as described above. According to the present example, the signal processing means 4 outputs an output signal to a wireless transmitter 7 that sends the converted signal to a remote subscriber. The present example is particularly suitable for handsfree telephony, e.g., in a vehicle.

Figure 3 shows steps of an embodiment of the herein disclosed method for reduction of non-acoustic perturbation, in particular, wind noise in microphone signals. Audio signals, e.g., in a vehicular cabin, are detected by one or more directional and one or more omni-directional microphones 10. For example and convenience of description only, it is assumed that one directional microphone and one omni-directional microphone are used.

After the frequency responses of the employed microphones have been adjusted to each other wind noise can be detected by means of comparing 11 the signal power of the microphone signals. If the signal power of the omnidirectional microphone is significantly lower by a predetermined value than the one of the directional microphone, it is decided that significant wind noise that deteriorates the microphone signal from the directional microphone is present.

Comparison is restricted to the low frequency range only, since wind noise represents low frequency perturbations.

If significant wind noise, i.e. wind noise above a predetermined level, is detected 12 the microphone signal from the omni-directional microphone is used to generate an output signal 13. If no significant wind noise is detected 12, the microphone signal from the directional microphone used to generate an output signal 14. According to this example in the case of absence of wind noise above the predetermined level, the microphone signal from the omnidirectional microphone is not used for the generation of the output signal and if wind noise above the predetermined level is present, the microphone signal from the directional microphone is not used for the generation of the output signal.

According to an advantageous embodiment of the inventive method processing of the digitized and Fourier transformed microphone signals is carried out in a time-dependent as well as frequency-dependent manner in the frequency domain. As shown in Figure 4 audio signals are detected 10 by both an omnidirectional microphone and a directional microphone. Again, multiple directional and omni-directional microphones may be employed. Depending on the level of wind noise a threshold frequency f _c (n) is determined 15 for a discrete time index n, below which non-acoustic perturbations, as wind noise, does occur. The threshold frequency f _G is time-dependently determined from the spectral short-time power of the microphone signals. To avoid artifacts caused by abrupt changes of the output signal, f _c is kept constant over some period, say, over one second.

For the frequency range above f _c the respective portion of the signal from the directional microphone is used to generate an output signal, since this frequency range is not affected by wind noise. Wind noise only affects the low frequency range, say, below 500 Hz to 1000 Hz. Thus, a combined signal comprising a portion of the microphone signal from the omni-directional microphone for frequencies below f _c and comprising a portion of the microphone

signal from the directional microphone for frequencies above f ₀ is generated 16. Then, the combined signal showing a significantly enhanced quality is output in the entire available frequency range 17.

All previously discussed embodiments are not intended as limitations but serve as examples illustrating features and advantages of the invention. It is to be understood that some or all of the above described features can also be combined in different ways.