EP3565276B1 - Method for operating a hearing aid and hearing aid - Google Patents

Description

The invention relates to a method for operating a hearing aid and a hearing aid which is set up in particular to carry out the method.

Hearing aids, in particular in the form of hearing aids, are used by people with a hearing loss to at least partially compensate for the hearing loss. For this purpose, conventional hearing aids regularly include at least one microphone for capturing noises from the environment and a signal processing processor, which is used to process the captured noises and, in particular, to amplify and / or attenuate them as a function of the individual hearing impairment (particularly frequency-specific). The processed microphone signals are forwarded by the signal processing processor to an output converter - usually in the form of a loudspeaker - for output to the hearing of the respective hearing aid wearer. Depending on the type of hearing loss, so-called bone conduction headphones or cochlear implants are also used as output transducers for mechanical or electrical stimulation of the hearing. The term hearing aid also includes other devices such as headphones, so-called tinnitus maskers or headsets.

Hearing aids in particular often have what is known as a classifier, which is used to infer certain, predefined “hearing situations”, in particular on the basis of the recorded noises. The signal processing is then changed as a function of the recognized hearing situation. Since the hearing aid wearer's speech understanding is frequently impaired due to the hearing impairment present, the (signal processing) algorithms stored in the signal processing processor are usually adapted to the utterances of third parties to work out in the recorded microphone signals and to reproduce them for the respective hearing aid wearer in a form that is as understandable as possible. To recognize a conversation situation, a speech recognition algorithm is often processed in the classifier. However, such an algorithm becomes imprecise in situations in which several people are speaking in the immediate vicinity of the hearing aid wearer, but not all of them are participating in the same conversation. In this case, acoustic identification of the people participating in the same conversation is regularly made more difficult.

the end EP 3 154 277 A1 a method for operating a hearing device is known in which the hearing device has two acceleration sensors which each provide a sensor signal and by means of which an orientation of the user's head is determined. The sensor signals from both acceleration sensors are used in combination to determine the orientation. An operating parameter of the hearing device is then set as a function of the orientation.

the end EP 2 928 210 A1 a binaural hearing aid system is known which comprises left and right hearing aid devices and a user interface, as well as its use and an operating method therefor. The left and right hearing aids include at least two input units for providing a time-frequency representation of an input signal in a number of frequency bands and a number of time instances, and a multiple input unit noise suppression system that includes a multi-channel beamformer filter unit that is operable with the at least two input units is coupled and configured to provide a beam-shaped signal. The binaural hearing assistance system is configured in such a way that a user can indicate a direction to or the location of a target signal source relative to the user via the user interface.

the end DE 10 2016 205 728 B3 a method for physically adapting a hearing aid to a hearing aid wearer is known. According to the method, a characteristic measure for a current actual wearing position of the hearing aid is determined by means of a position sensor of the hearing aid. Using the characteristic measure for the actual wearing position is then determined a deviation of the actual wearing position from a predetermined target wearing position. As a function of this deviation, an instruction is issued to the hearing aid wearer to adapt the receiver connection means as a function of the determined deviation.

NESTI, ALESSANDRO et al .: "Human discrimination of head-centered visualinertial yaw rotations", EXPEERIMENTAL BRAIN RESEARCH, Springer International, DE, Vol. 233, No. 12, August 30, 2015, pages 3553-3564, XP035868321, ISSN: 0014-4819, DOI 10.1007 / S00221-015-4426-2 describes an investigation into head-centered yaw rotations (0.5 Hz), which are presented either in the dark or in combination with visual cues (visual display of points of limited display duration). The participants differentiated a reference movement, which was repeated unchanged for each experiment, from a comparison movement, the top speed of which was iteratively adjusted in order to measure the participants' difference threshold, ie the smallest perceptible change in the stimulus intensity.

KESHNER EA et al .: "Characterizing head motion in three planes during combined visual and base of support disturbances in healthy and visually sensitive subjects", Gait & Posture, Elsevier, Amsterdam, NL, Vol. 28, No. 1, July 1 2008, pages 127-134, XP022686191, ISSN: 0966-6362, DOI: 10.10l6 / j.gaitpost.2007.11.003 describes an investigation into whether multiplanar movements in the environment lead to head instability, especially when the visual environment moves in planes that are orthogonal to a physical disturbance. For this purpose, the displacement of surfaces in the sagittal plane was combined with visual field disturbances in 12 healthy (29-31 years) and 3 visually sensitive (27-57 years) adults. Center of Pressure (COP), tip head angle, and RMS values of head movement were calculated and a three-dimensional model of joint movement was developed to study gross head movement in three planes. It was found that subjects who stood quietly in front of a visual scene that was translating in the sagittal plane produced a significantly greater (p <0.003) head movement when yawing than on a translational platform. However, if the platform was moved in the dark or with a visual scene rotating in a rolling motion, head movement orthogonal to the plane of platform movement was significantly increased (p <0.02). Visually sensitive subjects with no history of vestibular disorder produced large, delayed compensatory head movements. The orthogonal head movements were significantly greater in visually sensitive persons in the dark (p <0.05) and in the dark (p <0.01) than in healthy persons with a stationary scene. Aguilar M et al .: "Prediction of Pitch and Yaw Head Movements via Recurrent Neural Networks", IJCNN 2003, PROCEEDINGS OF THE INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS 2003, Portland, OR, July 20-24, 2003; [JOINT CONFERENCE ON NEURAL NETWORKS], New York, NY: IEEE, US, Vol. 4, July 20, 2003, pages 2813-2818, XP010652922, DOI: 10.1109 / IJCNN.2003.1224017, ISBN: 978-0-780, - 7898-8 describes that in virtual environment (VE) applications, in which virtual objects are presented in a head-mounted display, the virtual images must be continuously stabilized in space against the user's head movement. Latencies in head movement compensation lead to virtual objects floating around instead of being stable in space. This leads to an unnatural feeling, disorientation and simulation nausea, as well as errors in the adaptation of virtual and real objects. Visual update delays are a critical technical barrier to the implementation of head-mounted displays in a wide variety of applications. To address this problem, it is proposed to use machine learning techniques to define a forward model of head movement based on angular velocity information. In particular, a recurrent neural network is used to record the temporal pattern of the pitch and yaw movement. The results show the ability to move the head up to 40 ms.

The invention is based on the object of enabling improved operation of a hearing aid.

This object is achieved according to the invention by a method with the features of claim 1. Furthermore, this object is achieved according to the invention by a hearing aid with the features of claim 16. Advantageous and in part inventive embodiments and developments of the invention are in the subclaims and the following Description set out.

The method according to the invention is used to operate a hearing aid which (preferably only) has one acceleration sensor. This acceleration sensor is positioned on the head of a hearing aid wearer when it is worn as intended. Furthermore, the acceleration sensor is set up for measurement in at least two measurement axes that are perpendicular to one another (also referred to as “measurement directions”). According to the method, at least one main feature is derived from an acceleration signal of the acceleration sensor, which is related to an acceleration directed tangentially (and preferably approximately horizontally) to the head of the hearing aid wearer. The presence of a yaw movement of the head is then determined on the basis of the respective main feature (s), taking into account at least one predetermined criterion which can be derived from the acceleration signal itself and which goes beyond the presence of an acceleration value of the tangentially directed acceleration indicative of a movement.

“Related to the acceleration directed tangentially to the head of the hearing aid wearer” is understood here and below to mean that the main feature directly reproduces this tangentially directed acceleration, or that the main feature contains at least information about this.

Here and in the following, “yaw movement” is understood to mean, in particular, a rotational movement of the head about a vertical axis (which preferably coincides at least approximately with the vertical). Further terms for fundamental movements of the head are used here and in the following in particular "nodding" or "nodding movement" for an up and down movement directed up and down around a "nodding axis" that is preferably horizontal and in particular connecting the ears of the hearing aid wearer, as well as "rolling" or " Rolling motion "for a sideways inclination or tilting of the head around a "roll axis" which is preferably horizontal and in particular is oriented in the neutral viewing direction (also referred to as "zero-degree viewing direction").

"Acceleration sensor" is understood here and in the following in particular to mean a sensor in which sensor elements for measurement in the at least two measurement axes (ie for two-dimensional measurement), preferably in three measurement axes perpendicular to one another (three-dimensional measurement), are integrated. Such an acceleration sensor therefore preferably represents a self-contained component that is set up for connection to an evaluation unit.

According to the invention, a yaw movement is preferably not concluded when an acceleration directed tangentially to the head can be read from the acceleration signal, but only when it is concluded that the yaw movement is present, taking into account the at least one additional criterion. In this case, the probability that there is actually a yaw movement is increased. Misinterpretations of the acceleration signal can thus be avoided or at least reduced. Furthermore, it is advantageously made possible to use only one (single) acceleration sensor to detect the yaw movement, so that the use of conventionally used measuring systems that have multiple sensors, for example a combination of acceleration sensors with gyroscopes and / or magnetic field sensors (also known as "inertial measuring units") use, and the associated comparatively high energy consumption can be dispensed with. Furthermore, according to the invention, the recognized yaw movement is used to support the analysis of hearing situations.

In an expedient variant of the method, a time profile of the tangentially directed acceleration (also referred to as “tangential acceleration” for short in the following) is used as the main feature. In this case, the predefined criterion used, and thus considered, is whether the temporal course of the tangential acceleration is within a predefined movement time window has two oppositely directed local extrema consecutively (that is, for example, a local maximum and a local minimum). In particular, it is considered whether in the course of time the tangential acceleration assumes values with opposite signs at these two extremes. This is based on the knowledge that when the head is yawing, the tangential acceleration initially indicates an "actual" acceleration and then a "negative" acceleration (namely when the head is braked) with a respective associated deflection (the respective extreme) over time. In particular, depending on the alignment of the measuring axis assigned to the tangential direction relative to the actual direction of movement, the tangential acceleration thus initially assumes positive values, for example, and "changes" to negative values when the head is braked. When yawing the head in the opposite direction, the values of the tangential acceleration change accordingly from negative to positive. In this variant of the method, the movement time window is preferably adapted to the duration of a normal head turning movement, especially in a group conversation, and preferably has values between 0.25 and 2 or 1.5 seconds, in particular from 0.5 to 1 second. The movement time window is preferably "opened" (ie its monitoring is started) when a sufficiently significant change in the values of the tangential acceleration is detected. The movement time window advantageously limits the consideration of the main feature, in particular the temporal course of the tangential acceleration, so that "acceleration events" which, due to their comparatively long duration, are very likely not to be assigned to a head rotation (i.e. no yawing) , are not taken into account.

Here and in the following, an extremum of the temporal course is only inferred if the underlying change in the temporal course differs from a usual measured value fluctuation, e.g. noise, or from minor movements (which regularly do not cause sufficiently significant changes in the temporal course) can distinguish. For example, a threshold value comparison is carried out for this purpose.

In an expedient development of the method variant described above, an additional feature derived from the acceleration signal is a time curve of an acceleration directed radially (and in particular also horizontally) in particular with respect to a yaw axis of the head of the hearing aid wearer (which regularly at least approximately coincides with the vertical). In this case, the predefined criterion used is, in particular, whether the time profile of the radially directed acceleration (hereinafter referred to as “radial acceleration”) assumes a local extreme within the predefined (and above-described) movement time window. I. E. In this variant of the method, two criteria are considered, namely whether the tangential acceleration indicates the acceleration described above and the braking, and whether the radial acceleration also indicates an acceleration. From such an actually occurring radial acceleration, which is linked in particular to the centrifugal force which inevitably occurs during a yaw movement, a comparatively high probability can advantageously be derived - in particular in connection with the two local extremes of the tangential acceleration - that not just one straight-line movement is present along one of the measuring axes, but actually a yaw movement of the head.

In a method variant according to the invention - optionally in addition to the method variants described above - a movement intensity is determined on the basis of the temporal course of the tangential and possibly also the radial acceleration. In this case, a strength of the movement intensity - preferably the size of the determined value of the movement intensity - is used as a predetermined criterion. For example, a threshold value comparison is carried out for this purpose in order to compare the determined value of the movement intensity with a predetermined threshold value. In this case, the existence of the yaw movement is inferred in particular if the movement intensity has a specific, in particular predetermined, strength. Optionally, the presence of the yaw movement is not inferred if the movement intensity clearly exceeds and / or exceeds the specified (expected) strength falls below. In particular, in this variant of the method, a probability for the presence of the yaw movement is determined, the probability value of which decreases the further the movement intensity deviates from the expected strength (in particular it exceeds or falls below it).

A duration of movement and / or a total energy or mean energy contained in the tangential and radial acceleration - in particular in the respective measured value curve - is preferably determined as a measure of the movement intensity.

In a further variant of the method according to the invention, a correlation coefficient between a time derivative of the tangential acceleration and the radial acceleration is determined as the main feature. In particular, the tangential acceleration (preferably its time profile) is thus first derived from time and then correlated with the radial acceleration (preferably with its time profile). In this case, the strength - ie in particular the magnitude of the value - of the correlation coefficient is used as the predetermined criterion. For example, a threshold value comparison of the correlation coefficient with an in particular predetermined threshold value also takes place here. This approach is based on the knowledge that with a yaw movement of the head, the change in the tangential acceleration (i.e. the time derivative) - regularly - assumes a local extreme that closely coincides in time with the local extreme of the radial acceleration or even with it covered. This correlation coefficient thus advantageously represents an indicator that is comparatively easy to check for the presence of a yaw movement. A high magnitude value of the correlation coefficient can thus advantageously be used to infer a high probability of the presence of the yaw movement. A comparatively low strength of the correlation coefficient (for example less than 0.5 or less than 0.3), on the other hand, indicates comparatively uncoordinated or aimless head movements or an immobile head.

In an advantageous development of the method variant described above, a yaw direction is determined on the basis of the correlation coefficient, preferably on the basis of the sign. I. E. the sign of the determined correlation coefficient is used as an indicator for the direction in which the hearing aid wearer turns his head. This is due in particular to the fact that the acceleration sensor used has a positive and a negative measuring direction for each measuring axis. If, for example, the acceleration sensor is arranged on the left ear of the hearing aid wearer when the hearing aid is worn as intended and the measurement axis assigned to the tangential acceleration points with its positive direction in the viewing direction of the hearing aid wearer, the temporal course of the tangential acceleration in the case of a yaw movement to the left (despite the actual Acceleration) initially show negative values. Conversely, the time course will initially assume positive values in the case of a yaw movement to the right. Thus, in an optional variant of the method, only one main feature is sufficient to respond to the presence of the yaw movement and its yaw direction, in particular in a robust manner, i. H. with a comparatively low susceptibility to errors to be able to close.

In a further method variant according to the invention - in addition or as an alternative to the method variants described above - the main feature used is a curve of a diagram in which the tangential acceleration is plotted against the radial acceleration. I. E. this curve is first determined. In this case, the geometric shape of this curve in particular is used as the specified criterion. Such a curve advantageously already contains the information from the two measurement axes relevant for a yaw movement. As an option, there is no need to determine additional features.

In the case described above, it is particularly preferred to check as a predetermined criterion whether the curve of the diagram described above approximates an ellipsoidal shape. Since, as described above, in the case of a yaw movement, measured values that change over time are to be recorded for both the tangential and the radial acceleration, the temporally consecutive values are Measured values of the acceleration sensor in the diagram described above on a path curved in one direction. In the case of a conscious and "ideal" yaw movement of the head - that is, a uniform movement extending exactly in one of the two measuring axes assigned to the tangential and radial acceleration, in particular a horizontally arranged plane - the measured values of the acceleration sensor would be rounded, crescent-like Describe shape. Due to the mostly existing, anatomically determined "inclination" of the yaw plane of the head and other influences that lead to a mostly stationary offset (e.g. gravitation), the curve often has at least an oval, possibly "open" (ie start - and end point do not coincide) shape. In the case of an undirected head movement, on the other hand, the curve described above will have other shapes, for example a zigzag-like course (that is to say with alternating directions of curvature). Here and in the following, ellipsoid is understood to mean in particular that the curve has a shape that is curved and approximately closed in one direction of rotation (i.e., in particular with a slight offset compared to the curve length) or is at least composed of several curve sections that are curved in this way and optionally connect straight sections .

In a preferred development, the direction of rotation of the curve described above - which can in particular be read from the time sequence of the individual measured values - is used to determine the direction of yaw. Thus, in this case, too, in an optional variant of the method, only one main feature is sufficient to ensure that the yaw movement and its yaw direction are present in a particularly robust manner, ie. H. with a comparatively low susceptibility to errors.

In a preferred variant of the method, the respective main feature or the respective main feature as well as the possibly additionally determined additional feature are ascertained slidingly over a time window which overlaps with a subsequent, in particular similar, time window. The length of the respective time window is about 0.25 to 2 seconds, especially about 0.5 to 1.5 seconds. In this case, an overlap of the subsequent time window with the preceding time window of approximately 0.25-1, in particular up to 0.75 seconds, is preferably used. The length of the (respective) time window results from the knowledge that a usual, conscious yaw movement of the head lasts about 0.5 seconds to one second. In particular, in this variant of the method, two or three measured values assigned to the two or three measuring axes are output by the acceleration sensor at a frequency of approximately 10-60 Hertz, preferably approximately 15-20 Hertz. These groups of measured values (ie the respective two or three measured values) are in particular temporarily stored in a buffer memory which can hold eight of these groups of measured values. A so-called "update rate" of the buffer memory is preferably about two Hertz. This means that a continuous determination of the respective main feature or features can be dispensed with. For example, if no change in one of the measured values is detected within this time window, a determination of the respective main feature or features can be omitted. Computational effort can advantageously be saved as a result.

In a further preferred variant of the method, a value of a yaw angle is determined from the acceleration signal only when the presence of the yaw movement is detected in particular according to one or more of the method variants described above. On the one hand, this is useful to save computational effort. On the other hand, it can advantageously be avoided that in particular stationary influences or slowly changing disturbance variables (for example the earth's gravitational field, an inclined head posture or the like) affect the determination of the yaw angle. In order to determine the yaw angle, the tangential acceleration, in particular its temporal course, is preferably integrated (twice), it being recognized that the above-described influences or disturbance variables have a particularly strong effect due to the integration, especially in the case of comparatively long integration periods. Because the yaw angle is only determined when there is actually a yaw movement, the time length of the section of the time profile of the tangential acceleration to be integrated can be determined can be kept particularly short, so that the influences described above have only a slight effect and a drifting of the result can be avoided particularly effectively.

In a preferred variant of the method, constant and / or linear measured value components are filtered from the acceleration signal, in particular from the tangential and radial acceleration - optionally also from the integrated tangential acceleration (in particular its temporal progression), i.e. H. removed. For example, a high-pass filter is used in a simple but expedient variant. In a further simple variant, a temporal (in particular sliding) mean value of the measured values assigned to the respective measuring axis is subtracted from the individual measured values. In this way, stationary influences (for example gravitation) or influences which change only comparatively slowly and which are detected by the acceleration sensor can be removed or at least reduced. Additionally or alternatively, linear trends are removed from the measured values, in particular from the respective time courses or optionally from the integrated tangential acceleration, in that what is known as “detrending” is used in particular.

In principle, compensation in particular for gravity is preferably carried out, preferably in the "bare" acceleration signal, in particular in that the acceleration signal is fed to the high-pass filter. As a result, the influence of gravity can be reduced at least to a significant extent. In addition or as an alternative to the high-pass filtering, the pitch and roll angles of the head are determined in relation to the gravitational field. Using this angle, a so-called "direction cosine matrix" is then determined, by means of which the present measurement data contained in the acceleration signal (ie the measurement values assigned to the respective measurement axes) are transformed from a hearing aid wearer-specific coordinate system to the "global" earth-related coordinate system , especially rotated. After this coordinate transformation, the measurement data are cleared of the influence of the gravitational field - or at least the remnants of it remaining after the high-pass filtering - and then the measurement data are transferred to the original coordinate system (ie coordinate system related to the hearing aid wearer) transformed back. In this way, the influence of the gravitational field can advantageously be removed at least to a large extent.

Additionally or alternatively, the integrated tangential acceleration is optionally also cleared of such (stationary or slowly changing) influences, for example by means of "detrending". This variant is based on the consideration that, due to the comparatively short duration of a yaw movement, the remaining drift is comparatively low or is at least contained as an approximately constant or linear influence within the time window to be considered (which is mapped in particular in the buffer described above). In this way, the integrated tangential acceleration can be cleared of these (optionally remaining) constant and / or linear measured value components in a simple manner.

In an expedient variant of the method, a classification algorithm is applied to the respective main feature or features and possibly the additional feature in order to determine the presence or at least a probability of the presence of the yaw movement. I. E. the respective main feature or the respective main feature and, if applicable, the additional feature are fed to a classification algorithm which serves to carry out the consideration described above with regard to the fulfillment of the criteria assigned to the respective main feature (and possibly the additional feature). Optionally, the classification algorithm is also set up to determine not only the presence of the yaw movement but also the yaw direction (ie the direction of rotation or rotation when yawing the head), the duration and / or the strength of the yaw movement or at least the head movement. A “Gaussian mixed mode model”, a neural network, a “support vector machine” or the like is used as the classification algorithm. Optionally, a classifier that is often already present in a hearing aid (in which, in addition to the usual classification algorithms, the corresponding classification algorithm described above is preferably implemented) is used. The classifier and thus also the classification algorithm are preferably on trains the respective expression of the respective main or additional feature (ie the respective criterion) indicative of the presence of the yaw movement. Optionally, especially in the case of the neural network, the classifier is also modeled in a self-learning manner.

In a further expedient variant of the method, a spatial area of interest of the hearing aid wearer is determined on the basis of the yaw movement itself, but preferably on the basis of the determined values of the yaw angle covered during the yaw movement. I. E. It is observed over a predetermined period of time - which is again preferably a sliding period of time with a duration of, for example, 20 seconds to two minutes, in particular about 30 seconds to one minute - in which viewing directions, in particular starting from a zero degree line of sight the hearing aid wearer turns his head. This area of interest is preferably determined by statistically evaluating the yaw angles determined within the predetermined period of time (i.e. specifically the individual values) and, in particular, by creating a histogram. Since people - and therefore also the hearing aid wearer - usually turn their head to the current area of interest by turning their head (i.e. yawing the head), an area can be read from the statistical evaluation, for example the histogram of past yaw movements, by comparing a great interest of the hearing aid wearer is or at least was.

According to the invention, the information about the yaw movement of the head of the hearing aid wearer, in particular the spatial area of interest described above, is used to adapt a signal processing algorithm for a conversation situation. For example, it is possible to derive from the yaw movement, in particular from the histogram created therefrom, in which spatial viewing area the current main interest of the hearing aid wearer lies and thus also where potential conversation partners are. This information is particularly expediently used together with the information of an acoustic classifier, ie the information from the movement analysis described above (ie the determination of the presence the yaw movement) are combined with those of an acoustic analysis (ie the acoustic classifier), which is also referred to as "fusion". For example, the acoustic classifier is used to fundamentally determine the existence of a conversation situation and, if necessary, also to determine from which spatial directions relevant acoustic signals (usually speech signals emanating from third parties) hit the hearing aid and thus the hearing aid wearer. The information about the yaw movement of the head is preferably used in this case to further delimit the spatial area in which the interlocutors of the hearing aid wearer are most likely to be. This is particularly useful, for example, in the event that the hearing aid wearer is in an acoustically ambiguous conversation situation in which at least two conversations take place in parallel, but the hearing aid wearer only takes part in one of the two conversations. This occurs, for example, in restaurants, bars or the like, in particular when people on one side of the hearing aid wearer are talking to one another, but the hearing aid wearer only speaks to people in front of him or on his other side. In this case it can regularly happen that the acoustic classifier interprets all incoming speech signals as belonging to the conversation. The yaw movement of the head can thus be used to determine where the hearing aid wearer is actually looking, and from this it can be concluded which speech signals do not belong to the conversation with a comparatively high degree of probability.

In an advantageous variant of the method, the above-described zero-degree viewing direction of the hearing aid wearer is referenced in particular using a nodding movement of the head, a vertical movement of the hearing aid wearer and / or a forward movement of the hearing aid wearer (optionally detected by means of a separate "movement classifier"). Such movements that can be derived from the acceleration signal are used in particular to detect movements such as nodding, drinking, standing up, activities such as tying shoes, walking, jogging, driving a car, cycling and the like. This variant of the method, which also represents an independent invention, is based on the knowledge that, in particular, movements such as nodding and drinking are also possible during a group conversation or a lecture situation in which the hearing aid wearer If you look at a blackboard or a canvas for comparatively long periods of time, it is very likely that this will be done regularly with your head aligned in a zero-degree line of sight. The referencing serves to avoid or at least compensate for a drift, in particular when creating the above-described histogram, which may be caused, for example, by incorrect non-detection of a yaw movement. Furthermore, it can be assumed with a high probability that activities such as standing up and tying shoes are carried out with the head straight. The same applies to activities such as walking, jogging, driving a car, cycling and the like, in which the hearing aid wearer is very likely to turn his head to one side only relatively rarely. Optionally, a "movement classifier" is used to detect the movements described here, in particular activities such as walking, jogging, driving, cycling, tying shoes, in which the entire body of the hearing aid wearer is in particular in motion. This is preferably formed by a corresponding classification algorithm, which in turn is expediently directed to movements of the entire body of the hearing aid wearer.

In a further expedient variant of the method, an output of the above-described movement classifier, in particular aimed at the movement of the entire body of the hearing aid wearer, is used as an additional criterion for determining the yaw movement (in particular whether it is present). For example, it is assumed that in the case of activities recognized by means of the movement classifier, such as cycling, driving and jogging, the probability that the hearing aid wearer will take part in a group conversation is comparatively low. It is recognized that these activities each take place in comparatively “fast” movement situations in which the hearing aid wearer is likely to direct his (in particular visual) attention largely forwards with a comparatively high degree of probability. Using the information (output) of the movement classifier, the evaluation of the main features and, if applicable, the additional feature can be blocked or at least verified. If the hearing aid wearer is at rest, there will be multiple yaw movements of the head - especially in the case of a acoustically classified conversation situation - indicate with a high probability that the hearing aid wearer will participate in the conversation with several people. The information of the movement classifier can thus also be included, for example, in the classification algorithm described above (directed at the yaw movement) and / or in the amalgamation of the movement and acoustic information.

In a preferred variant of the method, such an arrangement of the acceleration sensor in or on the hearing aid is used that at least one of the measurement axes of the acceleration sensor is at least approximately tangential to the head, preferably parallel to the natural zero-degree viewing direction of the hearing aid wearer. This measuring axis is preferably also aligned horizontally. The two other measuring axes are preferably arranged (with an upright posture) vertically or horizontally and along the above-described pitch axis. As a result of this arrangement, the individual measured values assigned to the measuring axes are advantageously already assigned to the tangential and radial accelerations.

In particular in the event that the hearing aid described above is part of a binaural system, the above-described method for detecting the yaw movement and, if necessary, for determining the yaw angle is used separately in each of the two hearing aids. H. monaural - carried out and the two monaural decisions then "binaurally" synchronized.

In an optional variant of the method, the two monaural acceleration signals are combined to form a binaural signal - for example the difference is formed from the two acceleration signals - and the method described above is applied to the binaural sensor signal.

The hearing aid according to the invention comprises the (in particular single) acceleration sensor which, when the hearing aid is worn as intended, is arranged on the head of the hearing aid wearer and is set up for measurement in the at least two, optionally three, measuring axes. The hearing aid also includes a (signal processing) processor which - in terms of program and / or circuitry - is set up to carry out the above-described method according to the invention, in particular automatically. The processor is therefore set up to derive the at least one main feature linked to the tangential acceleration from the acceleration signal of the acceleration sensor and to determine the presence of the yaw movement of the head on the basis of the respective main feature, taking into account the at least one predetermined criterion. The hearing aid thus has all the advantages and features that result from the method features described above in equal measure.

In a preferred embodiment, the processor is at least essentially formed by a microcontroller with a microprocessor and a data memory in which the functionality for performing the method according to the invention is implemented in the form of operating software (firmware) so that the method - possibly in interaction with the Hearing aid wearer - is carried out automatically when the operating software is executed. The processor is alternatively a non-programmable electronic component, e.g. B. an ASIC, in which the functionality for performing the method according to the invention is implemented with circuitry means.

The conjunction “and / or” is to be understood here and in the following in particular in such a way that the features linked by means of this conjunction can be designed both together and as alternatives to one another.

Fig. 1

in einem schematischen Schaltbild ein Hörgerät,

Fig. 2

in einer Ansicht von oben einen Kopf eines Hörgeräteträgers mit dem bestimmungsgemäß am Ohr getragenen Hörgerät,

Fig. 3

in einem schematischen Ablaufdiagramm ein von einem Prozessor des Hörgeräts durchgeführtes Verfahren zum Betrieb des Hörgeräts,

Fig. 4, 5

jeweils in einem schematischen Diagramm aus einem Beschleunigungssignal abgeleitete Merkmale angetragen gegen die Zeit,

Fig. 6, 7

jeweils in einem schematischen Diagramm, in dem eine Radial- gegen eine Tangential-Beschleunigung angetragen ist, einen Verlauf der Beschleunigung,

Fig. 8

in einem schematischen Diagramm den zeitlichen Verlauf eines Gierwinkels des Kopfs des Hörgeräteträgers, und

Fig. 9

in einem schematischen Polardiagramm ein Histogramm des Gierwinkels.

The invention is explained in more detail below on the basis of exemplary embodiments. Show in it:

Fig. 1

a hearing aid in a schematic circuit diagram,

Fig. 2

in a view from above a head of a hearing aid wearer with the hearing aid intended to be worn on the ear,

Fig. 3

in a schematic flowchart a method for operating the hearing aid carried out by a processor of the hearing aid,

Fig. 4, 5

Characteristics derived from an acceleration signal plotted against time in a schematic diagram,

Fig. 6, 7

each in a schematic diagram in which a radial versus a tangential acceleration is plotted, a course of the acceleration,

Fig. 8

in a schematic diagram the time course of a yaw angle of the head of the hearing aid wearer, and

Fig. 9

in a schematic polar diagram, a histogram of the yaw angle.

Corresponding parts and sizes are always provided with the same reference symbols in all figures.

In Figure 1 a hearing aid 1, specifically a so-called behind-the-ear hearing aid, is shown. The hearing aid 1 comprises a (hearing aid) housing 2 in which several electronic components are arranged. The hearing aid 1 comprises two microphones 3 as electronic components, which are set up to detect noises from the surroundings of the hearing aid 1. Furthermore, the hearing aid 1 comprises a signal processor 4 as an electronic component, which is set up to process the noises detected by the microphones 3 and to output them to a loudspeaker 5 for output to the hearing of a hearing aid wearer. To detect the spatial position of the hearing device 1, it also includes an acceleration sensor 6 which is connected to the signal processor 4. In order to supply these electronic components with energy, a battery 7 is also arranged in the housing 2, which in the present exemplary embodiment is specifically formed by an accumulator. In order to conduct the sound generated by the loudspeaker 5 to the hearing of the hearing aid wearer, a sound tube 8 is connected to the housing 2 which, when worn as intended, is attached to the head 9, specifically to the ear of the hearing aid wearer (cf. Figure 2 ) is inserted with an ear mold 10 into the ear canal of the hearing aid wearer.

The acceleration sensor 6 is set up for three-dimensional measurement and has three measurement axes x, y and z that are perpendicular to one another (see Fig. Figure 2 ) on. The acceleration sensor 6 is arranged in the housing 2 of the hearing aid 1 in such a way that the measurement axis z coincides with the vertical direction when the head 9 is worn as intended and when the hearing aid wearer is in an upright posture. In this case, the measuring axis x is tangential to the head 9 and oriented to the front - that is, along a zero-degree viewing direction 12. The measurement axis y is directed radially away from the head 9. The two measuring axes x and y also lie in a horizontal plane when the hearing aid wearer is in an upright posture. Due to this arrangement, the measured values assigned to the measuring axis x reproduce an acceleration directed tangentially to the head 9 (hereinafter referred to as “tangential acceleration at”). The measured values assigned to the measuring axis y correspondingly reproduce an acceleration directed radially to the head 9 (hereinafter referred to as “radial acceleration ar”).

The signal processor 4 is set up to use an acoustic classifier, which is implemented as an algorithm in the signal processor 4, to infer a conversation situation (i.e. a conversation between at least two people) based on the noises recorded by the microphones 3 and then the signal processing accordingly adapt. For example, an opening angle of a directional microphone formed by means of the two microphones 3 is set in such a way that all speech components from the environment that hit the microphones 3, specifically the source locations of these speech components, are within the opening range of the directional microphone. In order to be able to adapt the signal processing even more precisely in such a conversation situation, specifically to be able to set the opening angle so that only the people actually involved in the conversation (who each represent a source of a part of the speech) are within the opening range of the directional microphone, the signal processor 4 carried out a method explained in more detail below.

In a first method step 20, the measured values determined by the acceleration sensor 6 - which are output in groups of three measured values, each of which is assigned to one of the measuring axes x, y and z - are output in a buffer memory (the one in the signal processor 4 is integrated) filed. The buffer memory is designed for the floating intermediate storage of eight such groups of measured values. In a subsequent method step 30, several features are derived (also: "extracted") from the measured values assigned to the respective measuring axes x, y and z. In a further method step 40, these features are fed to a classifier in which a classification algorithm - in the present exemplary embodiment in the form of a Gaussian mixed mode model - is implemented. Using the features derived in method step 30, this classifier determines whether the hearing aid wearer rotates his head 9, ie at least approximately rotates about the measurement axis z. Such a “sideways rotation” of the head 9 is referred to here and in the following as a “yaw movement”.

The measurement axis z represents a so-called yaw axis in the arrangement and alignment of the acceleration sensor 6 shown in the present exemplary embodiment which the hearing aid wearer tilts his head 9 downwards or upwards ("nod"; analogous to the English-language terms "yaw", "roll" and "pitch").

In parallel to the method steps 30 and 40 described above, in a method step 50 the measured values of the acceleration sensor 6 stored in the buffer memory are cleared of stationary influences that change only slowly compared to the duration of a head movement. The influence of gravity, which can be assumed to be stationary, is removed by means of a high-pass filter. Further influences that lead to an offset of the measured values, for example an anatomically-related deviation of the actual yaw axis from the vertical and / or the actual alignment of the measuring axis z, are in one embodiment by subtracting the time average of the buffered measured values from the respective "individual measured value " removed. Influences that have a linear effect (ie linear trends) are removed by means of so-called "detrending".

In the event that in a method step 55 the classifier outputs the result that there is a yaw movement of the head 9, in a further method step 60 a value of a yaw angle W is determined from the determined measured values, specifically from the tangential acceleration at. I. E. it is determined how far the hearing aid wearer has turned his head 9 (cf. Figure 8 ).

On the basis of the information as to whether there is a yaw movement and the yaw angle W by which the head 9 is rotated, a statistical analysis is carried out in a method step 70. It is determined how often the hearing aid wearer turns his head 9 within a predetermined time window. In addition, the values of the yaw angle W assigned to the individual yaw movements are used to create a histogram from which it can be read off in which directions - based on the zero-degree viewing direction 12 - the hearing aid wearer has turned his head 9 in the specified time window (see Fig. Figure 9 ). On the basis of the frequency distribution of the individual directions, a spatial distribution of the area of interest of the hearing aid wearer can also be read from this histogram.

In a further method step 80, the information generated in method steps 60 and 70 is used by the signal processor 4 in order to additionally adapt the signal processing. Specifically, in this method step 80, the information from the acoustic classifier described above and the “movement analysis” described above are merged by means of the acceleration sensor 6 in order to enable the signal processing to be more precisely adapted to a conversation situation. In one embodiment, the opening angle of the directional microphone, the alignment of the directional cone of the directional microphone and the position of a so-called "notch" (also referred to as a "notch") are specifically determined as a function of the information determined by the acceleration sensor 6 - namely the yaw angle W and the histogram - further adapted, possibly further limited with respect to a setting proposed solely by the acoustic classifier.

In a first exemplary embodiment, in method step 30, a time profile at (t) of the tangential acceleration at is determined as a main feature. As an additional feature, a time curve ar (t) of the radial acceleration ar is determined. In method step 40 it is considered as a criterion for the presence of the yaw movement whether the temporal course at (t) of the tangential acceleration at within a predetermined time segment, hereinafter referred to as "movement time window Zb", of a duration of one second, two local extremes Assumes Mt with opposite signs, indicating two opposite accelerations, namely an actual acceleration and a deceleration. Furthermore, within the scope of the criterion, it is also considered whether the time profile ar (t) of the radial acceleration ar assumes a local extreme Mr within the movement time window Zb, which indicates a head movement with an acceleration component directed radially towards the head 9. In Figure 4 the time courses at (t) and ar (t) are shown by way of example for a yaw movement of the head 9 to the right (cf. seconds 0.5-1.5) and to the left (cf. seconds 2-3). With the yaw movement to the right, the time course at (t) therefore runs through first the "positive" extremum Mt, which indicates the beginning of the yaw movement, and then the "negative" extremum Mt, which shows the Decelerating the head 9 indicates the end of the yaw movement. At the same time, the time course ar (t) - due to the alignment of the measurement axis y outwards - also shows a positive extreme Mr within the movement time window Zb due to the centrifugal force. Correspondingly, the opposite is true for the yaw movement to the left, as from the right half of Figure 4 can be taken. If such an expression of the main feature and the additional feature - ie as shown between seconds 0.5 and 1.5 or 2 and 3 - is recognized in method step 40, the classifier outputs in method step 55 that a yaw movement is present. Without the extreme Mr of the time profile ar (t), that is to say without any actual radial acceleration ar, there is, for example, only a straight forward movement of the head 9 or of the hearing aid wearer.

In a further exemplary embodiment, the main feature in method step 30 is a correlation coefficient K between a time derivative of the tangential acceleration at, specifically its time profile at (t), and the Radial acceleration ar, specifically its time course ar (t) is determined. This is in Figure 5 shown in more detail. The change in the tangential acceleration at, which can be taken from the time derivative of the tangential acceleration at, specifically a temporal extreme Md of this change, turns out to be as follows Figure 5 it can be seen - in the case of a yaw movement of the head 9, at least approximately temporally together with the extreme Mr of the radial acceleration ar. It can thus be inferred from the value of the correlation coefficient K - specifically from its magnitude - whether there is any yaw movement at all. The direction of the yaw movement can also be read from the sign of the correlation coefficient K. For the in Figure 5 yaw movement to the right shown between the seconds 0.5-1.5, the value of the correlation coefficient K is approximately -0.75. For the yaw movement to the left shown between seconds 2-3, the correlation coefficient K is approximately 0.8.

In a further exemplary embodiment, explained with reference to FIG Figures 6 and 7 In method step 30, a curve D of a diagram is created as the main feature in which the radial acceleration ar is plotted against the tangential acceleration at. In the subsequent method step 40, the shape of this curve D is used as the criterion. Specifically, it is considered whether the curve D can be approximated to the shape of an ellipse. In Figure 6 are also the previous ones Figures 4 or 5 underlying measured values for the yaw movement to the right and in Figure 7 applied to the left. The offset shown between the respective starting point and the end point (the latter marked by a triangle on its apex) is caused by an inclined posture of the head. As a result, the shape of curve D also deviates from the ideal circular shape and rather corresponds to an oval or an ellipse. If the curve D has such a shape, the classifier concludes in method step 40 that the yaw movement is present and outputs a corresponding result in method step 55.

In yet another exemplary embodiment (not shown in more detail), a movement intensity I is determined as the main feature in method step 30. This is contained in the tangential and radial acceleration Energy mapped. The movement intensity I is estimated on the basis of the averaged vector norms of the respective vector of the tangential and radial acceleration at or ar. For example, the energy is estimated by a temporally discrete sum of the vector length of the resulting vector of the tangential and radial acceleration at or ar.

In Figure 8 the time course of the values of the yaw angle W determined in method step 60 is shown by way of example.

In Figure 9 the histogram determined in method step 70 is shown as an example in the form of a polar diagram. From this it is possible to read in concrete terms on the basis of the radial length of the hatched areas how often or for how long the hearing aid wearer has turned his head 9 into a specific angular range. From this, in turn, a spatial area of interest can be derived, which is used in method step 80 to set the aperture angle of the directional microphone accordingly. In this specific example, the hearing aid wearer is talking to two people, one directly opposite and one offset to the left by about 20-25 degrees.

In an optional exemplary embodiment, in a method step 90 (see dashed illustration in Figure 3 ) a so-called movement classifier is used in order to infer a movement situation of the hearing aid wearer, ie a movement state of the entire body or an activity comprising this, on the basis of the features determined in method step 30. For example, in method step 90 it is determined whether the hearing aid wearer is at rest or, for example, is riding a bicycle. In the event that the hearing aid wearer is at rest, there is also a higher probability that the hearing aid wearer will take part in a conversation with several third parties. In the event that he rides a bicycle, the probability that he will take part in such a conversation is comparatively low. In this case, the determination of the yaw movement in method step 40 and the subsequent method steps 60-80 are optionally omitted.

In a further optional exemplary embodiment, in method step 55, the classifier also outputs the (temporal) duration of the yaw movement and optionally also the strength of the yaw movement, specifically the movement intensity I.

In a further exemplary embodiment, not shown in more detail, a “reset” takes place in a further method step; H. referencing the zero-degree viewing direction 12 whenever an almost pure nodding movement, which is indicative of drinking, for example. As a result, the histogram can be created particularly precisely and robustly, since - even if yaw movements are not recognized - the zero-degree viewing direction 12 can be "found" again and again, thus preventing the individual values of the yaw angle W from adding up and thus erroneously it is assumed that the zero-degree line of sight 12 changes.

The subject matter of the invention is not limited to the exemplary embodiments described above. Rather, further embodiments of the invention can be derived from the above description by a person skilled in the art. In particular, the individual features of the invention described on the basis of the various exemplary embodiments and their design variants can also be combined with one another in other ways. For example, in a further exemplary embodiment, in method step 40, all of the features described above, specifically the main features and the additional feature, are checked for compliance with the respective criterion. The subject matter of the invention is determined by the attached set of claims.

List of reference symbols

1

Hearing aid

2

casing

3

microphone

4th

Signal processor

5

speaker

6th

Accelerometer

7th

battery

8th

Sound tube

9

head

10

Earpiece

12th

Zero degree line of sight

20th

Process step

30th

Process step

40

Process step

50

Process step

55

Process step

60

Process step

70

Process step

80

Process step

at

Tangential acceleration

ar

Radial acceleration

at (t)

temporal course

ar (t)

temporal course

K

Correlation coefficient

D.

Curve

I.

Exercise intensity

Mt, Mr, Md

Extremum

W.

Yaw angle

E.g.

Movement time window

x, y, z

Measuring axis

Claims (16)

1. Method for operating a hearing device (1) having an acceleration sensor (6), which, in the intended wearing state, is positioned on the head (9) of a hearing device wearer and which is configured for measurement in two measurement axes (x, y, z), which are perpendicular to one another, wherein, according to the method

   - at least one main feature (at(t), I, K, D) is derived from an acceleration signal of the acceleration sensor (6), said feature being related to an acceleration (at) directed tangentially to the head (9),

   - the presence of a yaw movement of the head (9) is determined on the basis of the or the respective main feature (at(t), I, K, D), taking into account at least one predetermined criterion, which can be derived from the acceleration signal itself and which goes beyond the presence of an acceleration value of the tangentially directed acceleration (at) indicative for a movement,

   and wherein the detected yaw movement of the head (9) of the hearing device wearer is used to support an analysis of listening situations and/or is used to adapt a signal processing algorithm for a conversation situation,
   characterized in that
   as a main feature, a movement intensity (I) is determined on the basis of the time course of the tangentially and, optionally, the radially directed acceleration (at,ar), and wherein a strength of the movement intensity (I) is used as a predetermined criterion,
   and/or
   a correlation coefficient (K) between a time derivative of the tangentially directed acceleration (at) and the radially directed acceleration (ar) is determined as a main feature, and wherein a strength of the correlation coefficient (K) is used as a predetermined criterion,
   and/or
   a curve (D) of a diagram is used as the main feature, in which diagram the tangential acceleration (at) is plotted against the radial acceleration (ar), and the geometric shape of the curve (D) is used as a predetermined criterion.
2. Method according to claim 1,
   a time characteristic (at(t)) of the tangentially directed acceleration (at) is used as the main feature, and wherein it is used as a predetermined criterion whether the time characteristic (at(t)) of the tangentially directed acceleration (at) successively has two oppositely directed local extremes (Mt) within a predetermined movement time window (Zb).
3. Method according to claim 1,
   wherein a time characteristic (ar(t)) of an acceleration (ar) directed radially to the head (9) is derived as an additional feature from the acceleration signal, and wherein it is used as a predetermined criterion whether the time characteristic (ar(t)) of the radially directed acceleration (ar) assumes a local extreme (Mr) within a predetermined movement time window (Zb).
4. Method according to claim 1,
   wherein a movement duration and/or a total energy or mean energy contained in the tangentially and radially directed acceleration (at,ar) is determined as movement intensity (I).
5. Method according to claim 1,
   wherein a yaw direction is determined on the basis of the correlation coefficient (K), in particular on the basis of its arithmetic sign.
6. Method according to claim 1,
   where, as a predetermined criterion, it is checked whether the curve (D) of the diagram approximates an ellipsoidal shape.
7. Method according to claim 1 or 6,
   wherein the yaw direction is determined on the basis of the direction of rotation of the curve (D).
8. Method according to one of claims 1 to 7,
   wherein the or the respective main feature (at(t), I, K, D) and, optionally, the additional feature (ar(t)) are determined on a sliding basis over a time window overlapping with a subsequent time window.
9. Method according to one of claims 1 to 8,
   wherein a value of a yaw angle (W) is determined from the acceleration signal only when the presence of the yaw movement is detected.
10. Method according to one of claims 1 to 9,
    wherein constant and/or linear measured value components are filtered out of the acceleration signal.
11. Method according to one of claims 1 to 10,
    wherein a classification algorithm is applied to the or the respective main feature (at(t), I, K, D) and, optionally, the additional feature (ar(t)) to determine the presence or at least a probability of the presence of the yaw movement.
12. Method according to one of claims 1 to 11,
    wherein a spatial area of interest of the hearing device wearer is determined over a predetermined period of time based on the yaw movement.
13. Method according to one of claims 1 to 12,
    wherein the information about the yaw movement of the head (9) of the hearing device user is used to adapt a signal processing algorithm for a group conversation situation.
14. Method according to one of claims 1 to 13,
    wherein a zero degree viewing direction (12) of the hearing device wearer is referenced based on a nodding movement of the head (9), based on a vertical movement of the hearing device wearer and/or based on a forward movement of the hearing device wearer.
15. Method according to one of claims 1 to 14,
    wherein such an arrangement of the acceleration sensor (6) in or on the hearing device (1) is used that one of the measuring axes (x,y,z) of the acceleration sensor (6) is aligned at least approximately tangentially to the head (9).
16. Hearing device (1),
    with an acceleration sensor (6) which is positioned on the head (9) of a hearing device wearer in the intended wearing state and which is configured for measurement in two measurement axes (x,y,z) which are perpendicular to one another, and having a processor (4) which is configured to carry out the method according to one of claims 1 to 15.

Images