Method, device and system for identifying signal characteristic of sensor signal of sensor

Abstract

The invention relates to a method and a device for identifying a signal characteristic of a sensor signal from a sensor (2), the method comprising the following steps: sampling (S1) the sensor signal (SIG) at a sampling rate to produce a sampled value (x); calculating (S2) a correlation metric (d1) for the sensor signal (SIG) by summing an absolute difference between two successive sampling values (xi + 1; xi) of a sampling sequence of the sampled sensor signal, the sampling sequence comprising a determined number N of sampling values; and determining (S3) a signal characteristic (SC) of the sensor signal (SIG) on the basis of the calculated correlation measure (d1). The invention also relates to a system.

Description

Technical Field

[0001] This invention proposes a method and apparatus for identifying signal characteristics of sensor signals (especially signals from MEMS sensors) from sensors, and in particular, a method and apparatus for extending the lifespan of sensors by switching to a special operating mode after a sensor fault is identified. Background Technology

[0002] EP2399099A1 describes a MEMS gyroscope with an adjustable drive frequency.

[0003] In microelectromechanical systems (MEMS), leakage current is typically generated when particles detach from a moving body and adhere to the electrodes below it, causing short circuits on the electrodes. These electrode short circuits manifest as strong noise in the sensor's output characteristics. The impact energy on the sensor plays a significant role in particle detachment. High impact energy can lead to the detachment of polycrystalline silicon particles of varying sizes and distributions, as is the case in MEMS inertial sensors. Debris, possibly several micrometers in size, from the suspended mass may still adhere to the block. However, due to the high impact energy, this debris can also reach locations several micrometers from the point of impact. Kinetic impact energies up to 40 nanojoules (velocities of 1–3 m / s for most inertial sensor masses) can be associated with silicon fracture and further particle detachment, resulting in leakage current and high noise. Traditionally, to avoid detached particles, modifications to the design of different buffer shapes in stationary / movable components can be considered. Different buffer geometries, such as pointed, planar, or circular, can be used for this purpose. Sharp protrusions result in less detachment compared to round or flat protrusions because, upon impact, the test mass is deflected by the sharp geometry, and the energy balance between the mass, suspension spring, and protrusion is achieved in a different way. Kinetic energy is transferred more to the lateral deflection of the spring than to the impact / mass deformation (and potential failure). While in principle, damper design can improve the robustness of inertial sensors to strong impacts, practical devices typically have a distributed damper arrangement rather than a single damper.

[0004] Another traditional approach to avoid noise without requiring design modifications to the MEMS sensor is to use positive electromechanical feedback (EMF) at the pull-in point, which is essentially defined by a voltage at which the two capacitor plates of the MEMS structure pull together due to electrostatic attraction. However, it is difficult to achieve the inherent measurement resolution of capacitive MEMS sensors using standard integrated electronics. This is often due to a mismatch between the MEMS structural elements and the readout electronics. This mismatch arises because standard readout devices typically prefer small or medium source impedances, while MEMS structural elements have high impedances. Furthermore, CMOS circuits are known to contribute a large 1 / f noise. One possibility for reducing impedance mismatch and 1 / f noise is to use high-frequency readout techniques (HF-Auslesetechniken), for which a small-signal model is used in noise and stability analysis.

[0005] The capacitance of a MEMS sensor (described by the moving mass m and the spring constant k) can be monitored using an RF-Kapazitätsbrücke (RF-Clock Bridge). Here, the output signal of the capacitor bridge is down-mixed, amplified, and fed back to the MEMS device. Since the carrier frequency is much higher than the mechanical band of the MEMS, the equivalent noise of any readout amplifier at the latch point can be completely eliminated. At the latch point, the contribution of readout noise increases with frequency, which limits the noise matching bandwidth. Therefore, particle shedding can cause leakage current, which can be reduced by hardware modifications (e.g., at the shape of the buffer) or by 1 / f noise suppression. However, traditional methods requiring hardware modifications necessitate complex noise suppression circuitry.

[0006] Traditional noise identification methods typically correlate a reference signal with a relevant signal. However, in many cases, such a reference signal is unavailable, necessitating its generation by a dedicated reference signal generator, thus increasing circuit complexity. For noise identification, traditional methods are generally associated with high computational demands, either involving more processing steps or requiring comparisons between the signal of interest and the reference signal. Summary of the Invention

[0007] According to a first aspect, the present invention proposes a method for identifying signal characteristics of a sensor signal from a sensor, the method comprising the steps of: sampling the sensor signal at a sampling rate to generate sampled values; calculating a correlation measure (Korrelationsmaßes) for the sensor signal by summing the absolute differences between two successive sampled values ​​of a sampling sequence of the sampled sensor signal, wherein the sampling sequence comprises a determined number of sampled values; and determining the signal characteristics of the sensor signal based on the calculated correlation measure.

[0008] A significant advantage of the method according to the invention is that it only requires processing the sensor signal from the sensor, without the need for a separate reference signal.

[0009] This represents a significant simplification in circuit technology and a reduction in complexity in signal processing. Furthermore, application possibilities are expanded.

[0010] The method according to the present invention does not require modification of existing circuits and is preferably based on corresponding matching software.

[0011] The method according to the present invention provides an efficient possibility for identifying noise in sensor signals.

[0012] Therefore, the method according to the invention is applicable to all sensors that provide analog sensor signals, and is particularly applicable to MEMS sensors.

[0013] Monitoring of sensor signals using the method according to the invention can be performed continuously in the background. Data processing of the monitored sensor signals is preferably performed in real time.

[0014] In one possible implementation of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, the calculated correlation metric is compared with a threshold.

[0015] In one possible implementation of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, the sensor signal is classified as noise if the calculated correlation metric is higher than a threshold.

[0016] The threshold can be predefined. Alternatively, the threshold can be set via an interface.

[0017] In one possible implementation of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, the sensor signal is classified as regular if the calculated correlation metric is below a threshold.

[0018] In one possible implementation of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, if the calculated correlation metric is zero, the sensor signal is classified as constant.

[0019] In one possible implementation of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, sampled values ​​of a sampled sequence are temporarily stored, and a correlation metric is calculated based on the temporarily stored sampled values ​​of the sampled sequence.

[0020] In one possible implementation of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, a correlation metric is calculated iteratively without temporarily storing the sampled values ​​of the sampled sequence. This approach enables lower latency.

[0021] In one possible implementation of the method for identifying signal characteristics of a sensor signal from a sensor according to the invention, another correlation measure for the sensor signal is calculated by summing the products of two consecutive sampled values ​​of a sampled sequence of the sampled sensor signal, wherein the sampled sequence comprises a defined number of sampled values.

[0022] The number N of sampled values ​​in the sampling sequence can be configured fixedly. Alternatively, the number N of sampled values ​​in the sampling sequence can be adjusted according to the specific application. The number N of sampled values ​​can also be dynamically adjusted during continuous operation of the sensor. For example, the number N of sampled values ​​in the temporarily stored sampling sequence can be adjusted based on a tolerable delay time.

[0023] In one possible embodiment of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, two correlation measures (d1, c1) calculated for the sensor signal are stored in pairs. In this way, the signal characteristics of the sensor signal can be determined more specifically.

[0024] In one possible embodiment of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, the ratio between two correlation measures (d1, c1) is calculated. In this way, the signal characteristics of the sensor signal can be determined more specifically.

[0025] In one possible embodiment of the method for identifying signal characteristics of sensor signals from a sensor according to the present invention, the signal characteristics of the sensor signal are further described in detail based on the calculated ratio of two correlation measures (d1, c1).

[0026] In one possible implementation of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, the sensor is evaluated as either fault-free or faulty based on the obtained signal characteristics of the sensor signal.

[0027] In one possible embodiment of the method for identifying signal characteristics of sensor signals from a sensor according to the present invention, an operating mode for handling the identified fault is automatically activated after a sensor fault is identified.

[0028] The handling of identified sensor malfunctions can vary depending on the application. The sensor may be disabled and / or its signal may be ignored. Alternatively, switching can be performed from another redundant sensor of the same type. Furthermore, error or warning messages can be output to the higher-level control unit of the auxiliary system, and / or a specially configured operating mode can be activated.

[0029] In one possible embodiment of the method according to the invention for identifying signal characteristics of sensor signals from a sensor, the sensor signal is output by a MEMS sensor of a gyroscope.

[0030] According to another aspect, the present invention provides a program for identifying signal characteristics of sensor signals from a sensor, the program comprising program instructions for executing the method according to the invention. Alternatively, the method steps according to the invention can also be implemented by hardwired hardware.

[0031] Furthermore, the present invention also proposes an apparatus for identifying signal characteristics of sensor signals from a sensor, the apparatus comprising: a sampling unit that samples the sensor signal at a sampling rate to generate sampled values; a calculation unit that calculates a correlation measure for the sensor signal by summing the absolute differences between two successive sampled values ​​of a sampling sequence of the sampled sensor signal, wherein the sampling sequence includes a determined number of sampled values; and a determination unit that determines the signal characteristics of the sensor signal based on the calculated correlation measure.

[0032] In one possible embodiment of the apparatus according to the invention for identifying signal characteristics of sensor signals from a sensor, the determining unit has a comparator that compares a correlation metric calculated by the calculation unit with an adjustable or fixed threshold, wherein if the calculated correlation metric is higher than the threshold, the determining unit classifies the sensor signal as noisy; if the calculated correlation metric is lower than the threshold, the determining unit classifies the sensor signal as regular; and if the calculated correlation metric is zero, the determining unit classifies the sensor signal as constant.

[0033] In one possible embodiment of the apparatus according to the invention for identifying signal characteristics of sensor signals from a sensor, the apparatus has a buffer memory for temporarily storing sampled values ​​of a sampling sequence generated by a sampling unit.

[0034] In one possible embodiment of the apparatus for identifying signal characteristics of a sensor signal from a sensor according to the invention, the computing unit of the apparatus calculates another correlation measure for the sensor signal by summing the product of two consecutive sampled values ​​of a sampled sequence of the sampled sensor signal, wherein the sampled sequence comprises a defined number of sampled values.

[0035] In one possible embodiment of the apparatus according to the invention for identifying signal characteristics of sensor signals from a sensor, two correlation metrics (d1, c1) calculated by the computing unit for the sensor signal are stored in pairs in the data memory of the apparatus.

[0036] In one possible embodiment of the apparatus according to the invention for identifying signal characteristics of sensor signals from a sensor, the computing unit of the apparatus calculates the ratio between two correlation measures.

[0037] In one possible embodiment of the apparatus for identifying signal characteristics of sensor signals from a sensor according to the present invention, the determining unit of the apparatus further details the signal characteristics of the sensor signal based on the ratio between two correlation metrics (d1, c1) calculated by the computing unit.

[0038] In one possible embodiment of the apparatus for identifying signal characteristics of sensor signals from a sensor according to the present invention, the evaluation unit of the apparatus evaluates whether the sensor is operating normally without faults or has a fault based on the determined signal characteristics of the sensor signal.

[0039] In one possible embodiment of the device for identifying signal characteristics of sensor signals from a sensor according to the present invention, if a sensor fault is identified based on the determined signal characteristics, the evaluation unit outputs an activation signal for activating an operating mode for handling the sensor fault.

[0040] Furthermore, the present invention proposes a system comprising at least one sensor and means for identifying signal characteristics of a sensor signal from the sensor, the means comprising: a sampling unit that samples the sensor signal at a sampling rate to generate sampled values; a calculation unit that calculates a correlation measure for the sensor signal by summing the absolute differences between two successive sampled values ​​of a sampling sequence of the sampled sensor signal, wherein the sampling sequence comprises a determined number of sampled values; and a determination unit that determines the signal characteristics of the sensor signal based on the calculated correlation measure.

[0041] In one possible implementation of this system, the sensor is a MEMS sensor. Attached Figure Description

[0042] Possible embodiments of the method and apparatus according to the invention will now be described in more detail with reference to the accompanying drawings.

[0043] The attached diagram shows: Figure 1 : A flowchart illustrating one possible implementation of the method according to the present invention; Figure 2 : A block diagram illustrating one possible embodiment of the device according to the invention; Figure 3 : Another flowchart illustrating one possible implementation of the method according to the invention; Figure 4A , Figure 4B : A signal diagram used to illustrate the working principle of the method and the device according to the present invention; Figure 5 : A graph used to show the correlation measure calculated for signal characterization; Figure 6 : A graph used to illustrate another correlation metric calculated for signal characterization; Figure 7 : A graph used to show the correlation metric calculated for signal characterization.

[0044] The accompanying drawings are intended to aid in a further understanding of embodiments of the invention. They illustrate embodiments and, in conjunction with the description, serve to explain the principles and concepts of the invention. Other embodiments and many of the advantages described above become apparent from the accompanying drawings. Elements in the drawings are not necessarily drawn to scale.

[0045] In the accompanying drawings, unless otherwise specified, identical elements, features, and components that are identical in function and effect are given the same reference numerals. Detailed Implementation

[0046] like Figure 1 As shown, the method for identifying signal characteristics of sensor signal SIG from sensor 2 according to the present invention includes several main steps.

[0047] In the first step S1, the sensor signal SIG is sampled at a sampling rate R to generate a sampled value (x). The sampling rate R can be preset or set by an internal control device.

[0048] In another step S2, a correlation measure (d1) for the sensor signal SIG is calculated by summing the absolute differences between two consecutive sampled values ​​(xi+1; xi) of the sampled sequence of the sensor signal SIG sampled in step S1, wherein the sampled sequence comprises a determined number of N sampled values.

[0049] In another step S3, the signal characteristics (SC) of the sensor signal SIG are determined based on the correlation metric (d1) calculated in step S2.

[0050] In one possible implementation of the method for identifying signal characteristics SC of sensor signal SIG from sensor 2 according to the present invention, such as according to Figure 3 As shown in the flowchart, if the calculated correlation metric (d1) is higher than an adjustable or predefined threshold (SW), the sensor signal is classified as noise (SC1) in step S3.

[0051] In one possible implementation of the method for identifying signal characteristics SC of a sensor signal from sensor 2 according to the present invention, if the calculated correlation metric (d1) is below a threshold (SW), the sensor signal SIG is classified as regular (SC2).

[0052] In one possible implementation of the method according to the invention for identifying signal characteristics SC of a sensor signal from sensor 2, if the calculated correlation metric (d1) is zero, the sensor signal SIG is classified as constant (SC3).

[0053] In one possible embodiment of the method for identifying signal characteristics of a sensor signal from sensor 2 according to the present invention, the sampled values ​​of the sampled sequence are temporarily stored, and a correlation metric (d1) is calculated in step S2 based on the sampled values ​​of the sampled sequence temporarily stored in the buffer memory.

[0054] In another possible implementation of the method for identifying signal characteristics of sensor signal SIG from sensor 2 according to the present invention, in step S2, the correlation metric d1 is calculated iteratively without temporarily storing the sampled values ​​of the sampled sequence.

[0055] It also allows switching between iterative and non-iterative computation. For example, in time-critical applications requiring low latency, iterative computation of the correlation metric d1 can be activated.

[0056] In one possible implementation of the method for identifying signal characteristics SC of a sensor signal SIG from sensor 2 according to the present invention, in step S3, another correlation measure (c1) for the sensor signal SIG is calculated by summing the products of two consecutive sampled values ​​(xi+1; xi) of the sampled sequence of the sampled sensor signal, wherein the sampled sequence comprises a determined number of N sampled values.

[0057] In another possible embodiment of the method according to the invention for identifying signal characteristics SC of sensor signal SIG from sensor 2, the two correlation measures (d1, c1) calculated for sensor signal SIG in step S3 are stored in pairs.

[0058] In one possible implementation of the method according to the invention for identifying signal characteristics SC of sensor signal SIG from sensor 2, in step S3, the ratio v between two correlation measures (d1, c1) is also calculated.

[0059] In one possible embodiment of the method for identifying signal characteristics of a sensor signal SIG from sensor 2 according to the present invention, the signal characteristics SC of the sensor signal SIG are further described in detail based on the ratio v of the two correlation measures (d1, c1) calculated in step S3. At this time, it is preferable to further describe the signal characteristics in detail based on the ratio between the two correlation measures, with regard to a determined value of one of the two correlation measures.

[0060] In one possible embodiment of the method for identifying signal characteristics of a sensor signal from sensor 2 according to the present invention, the sensor 2 is evaluated as either operating without faults or having a fault, based on the signal characteristics SC of the obtained sensor signal SIG. Upon identification of a fault in sensor 2, an operating mode for handling the identified fault can be automatically activated.

[0061] In one possible embodiment of the method according to the invention for identifying the signal characteristic SC of a sensor signal from sensor 2, the sensor signal SIG is output by the MEMS sensor 2 of the gyroscope. However, this method is applicable to a variety of other sensors that output analog sensor signals.

[0062] Figure 2An embodiment of a device 1 according to the present invention for identifying signal characteristics (SC) of a sensor signal SIG from a sensor 2 is shown. The device includes a sampling unit 1A that samples the sensor signal SIG at a sampling rate R to generate sampled values ​​(x). The sampling rate R can be predefined for the sensor 2 to be monitored. The sampling rate can also be adjusted by a local control device of the device 1 according to a configuration (e.g., according to the sensor type and / or application). In one possible implementation, the sampling unit 1A may be integrated into the sensor 2.

[0063] The device 1 also includes a computing unit 1B, which calculates a correlation measure (d1) for the sensor signal SIG by summing the absolute differences between two successive sampled values ​​(xi+1; xi) of a sampled sequence of the sampled sensor signal SIG, wherein the sampled sequence comprises a defined number N sampled values. The computing unit 1B may have a processor, a field-programmable gate array (FPGA), or an application-specific integrated circuit (ASIC).

[0064] The apparatus 1 further includes a determining unit 1C, which determines the signal characteristics SC of the sensor signal SIG based on the correlation metric d1 calculated by the calculation unit 1B. In one possible embodiment of the apparatus 1 for identifying the signal characteristics SC of the sensor signal SIG from the sensor 2 according to the invention, the determining unit 1C has a comparator that compares the correlation metric (d1) calculated by the calculation unit 1B with a threshold SW. Here, if the calculated correlation metric (d1) is higher than the threshold SW, the determining unit 1C classifies the sensor signal SIG as noise. If the calculated correlation metric (d1) is lower than the threshold SW, the determining unit 1C classifies the sensor signal SIG as regular. If the correlation metric (d1) calculated by the calculation unit 1B is zero, the determining unit 1C classifies the sensor signal SIG as constant.

[0065] In one possible embodiment of the apparatus according to the invention for identifying signal characteristics of a sensor signal SIG from sensor 2, apparatus 1 further includes a buffer memory for temporarily storing sampled values ​​of a sampling sequence generated by sampling unit 1A. The sampling sequence has a defined number N of sampled values. This number N can be fixedly defined. Alternatively, the number N of sampled values ​​is also dynamically matched during operation of sensor 2, for example, in response to an identified event or according to a maximum delay time required by apparatus 1.

[0066] In one possible embodiment of the apparatus according to the invention for identifying signal characteristics SC of a sensor signal SIG from a sensor 2, the computing unit 1B of the apparatus 1 calculates another correlation measure (c1) for the sensor signal SIG by summing the products of two consecutive sampled values ​​(xi+1; xi) of a sampled sequence of the sampled sensor signal, wherein the sampled sequence comprises a determined number of N sampled values.

[0067] In one possible embodiment of the apparatus according to the invention for identifying signal characteristics SC of sensor signal SIG from sensor 2, two correlation metrics (d1, c1) calculated by computing unit 1B for sensor signal SIG are stored in pairs in the data memory of apparatus 1.

[0068] The calculation unit 1B of device 1 can calculate the ratio v between two correlation measures (d1, c1) by performing a division operation (v = d1 / c1). In one possible embodiment of device 1, the signal characteristics SC of the sensor signal SIG can be further described in detail based on the ratio v between the two correlation measures (d1, c1) calculated by the calculation unit 1B, with respect to the value of one of the two correlation measures.

[0069] In one possible embodiment of the apparatus according to the invention for identifying signal characteristics of a sensor signal SIG from a sensor 2, the evaluation unit of the apparatus evaluates whether the sensor 2 is operating normally without faults or has a fault based on the determined signal characteristics SC of the sensor signal SIG. If a fault is identified in the sensor 2 based on the determined signal characteristics SC, the evaluation unit may output an activation signal for activating an operating mode for handling the fault in the sensor 2.

[0070] Furthermore, the present invention proposes a system SYS, which includes at least one sensor 2 and a device 1 for identifying signal characteristics SC of a sensor signal SIG from the sensor 2. The device includes: a sampling unit 1A that samples the sensor signal SIG at a sampling rate R to generate sampled values ​​(x); a calculation unit 1B that calculates a correlation measure (d1) for the sensor signal SIG by summing the absolute differences between two consecutive sampled values ​​(xi+1; xi) of a sampling sequence of the sampled sensor signal SIG, wherein the sampling sequence includes a determined number N sampled values; and a determination unit 1C that determines the signal characteristics SC of the sensor signal SIG based on the calculated correlation measure d1. Figure 2In one possible implementation of the system SYS shown, sensor 2 is a MEMS sensor. In another possible implementation, device 1 may also be integrated into the housing of the sensor 2 to be monitored.

[0071] In one possible implementation, it is further determined whether Enhanced Coverage (EC) mode should be enabled to achieve "high robustness" of signal sensor 2. To determine whether EC mode should be triggered, the signal SIG of sensor 2 (e.g., a gyroscope) is analyzed. Gyroscope signals have specific characteristics that indicate EC mode should be enabled. These signal characteristics are identified by an algorithm corresponding to the method according to the invention to automate the triggering of EC mode.

[0072] Essentially, two different signal characteristics SC can be distinguished from normal gyroscope signals. The main characteristic of the first signal characteristic SC1 is strong noise. This noise is broadband and can be considered as white noise.

[0073] The main characteristic of the second signal feature SC2 is that the signal from one of the three gyroscope axes is fixed or constant.

[0074] Typically, a gyroscope provides three sensor signals SIG, which measure the angular velocity about three orthogonal axes. These axes are usually called the x-axis, y-axis, and z-axis. Each of these SIG signals represents the rate of rotation about one of the three axes. The three main signals provided by a gyroscope include: Angular velocity about the x-axis: This measures rotational motion (roll) about the x-axis. Angular velocity about the y-axis: This measures rotational motion (pitch) about the y-axis. Angular velocity about the z-axis: This measures the rotational motion (yaw) about the z-axis.

[0075] The three sensor signals SC1, SC2, and SC3 of a gyroscope together enable a complete characterization of the rotational motion of an object (on which the gyroscope is mounted). Modern gyroscopes (such as those used in smartphones, drones, and vehicles) continuously provide these three sensor signals, which can then be used for control, navigation, or stabilization of the corresponding system. It is important to distinguish the SC1 signal with anomalous characteristics (especially the first signal characteristic SC1 (noise) and the second signal characteristic SC2 (fixed / constant signal)) from normal gyroscope signals.

[0076] In one possible implementation, a device 1 is set for each of the three sensor signals SIG of the gyroscope to identify the signal characteristic SC of the sensor signal SIG. Alternatively, the sensor signals SIG can be sequentially switched to the signal input of the computing unit 1B after sampling using a multiplexer. Once a fault is identified in one of the three sensor signals, the fault is automatically and appropriately handled.

[0077] Because the noise is high in the first signal characteristic (SC1), the variance VAR of the gyroscope signal can be used as a possible criterion for differentiation. However, a drawback of using variance VAR is that it does not contain information about the correlation of subsequent samples of the signal. But this information is important here, as this differentiation should also function correctly when sensor 2 is in motion.

[0078] In such cases, a third signal feature (SC3) can be considered, which also has a high variance VAR, such as a signal SS in the form of a sine curve. However, in these cases, the algorithm should not classify it as the first signal feature SC1, but rather as the third signal feature SC3.

[0079] In the method according to the invention, the following correlation measure d1 is used: this correlation measure includes the correlation between successive samples within a time range of N samples. In the presence of a sinusoidal signal SS, subsequent sampled values ​​are strongly correlated. However, a noisy signal RS (e.g., white noise) results in a low correlation value d1 between successive sampled values ​​because, ideally, each sampled value of the noisy signal RS is statistically independent and therefore uncorrelated.

[0080] exist Figure 4A , Figure 4B The diagram above shows two signals, SS and RS. Figure 4A The figure below shows a sinusoidal signal SS with a frequency of 100Hz. Figure 4B The random white noise signal RS is shown in the figure. Time t (in seconds) is plotted on the x-axis, and the signal values ​​(in arbitrary units) are plotted on the y-axis. Figure 4A , Figure 4B The sampling rate R of both signals shown is 6.4 kHz.

[0081] Figure 4A , Figure 4B The signals in the dataset all have a variance VAR of approximately 1, therefore it is impossible to distinguish between the two signals SS and RS based on the variance VAR. If we calculate... Figure 4A , Figure 4BThe correlation between successive samples of the two signals SS and RS is 6.4 × 10⁻⁶. 4 And -204. Therefore, the absolute values ​​differ by more than 312 times, making them clearly distinguishable. Figure 4A The sinusoidal signal SS shown is Figure 4B The noise signal RS shown is shown in the figure.

[0082] The correlation value c1 between successive samples or sampled values ​​for a frame of length N is given in the following formula (1): Where xi is the i-th sampled value of the sensor signal SIG.

[0083] However, using the correlation function in formula (1) has drawbacks in classifying the signal features, especially when identifying the second signal feature SC2. The main characteristic of the second signal feature SC2 is that the signal from one of the three gyroscope axes is fixed or constant. The correlation between successive sampled values ​​must be maximized because, in the case of the second signal feature SC2, successive sampled values ​​are indistinguishable from each other. This raises the question of defining the maximum value.

[0084] When calculating formula (1), it is unknown whether the calculated value is the maximum value of the autocorrelation function. Therefore, it is also necessary to calculate the following formula (2) to obtain the reference point (Anhaltspunkt). However, this requires additional calculations.

[0085] .

[0086] On the other hand, identifying the first signal characteristic SC1 is also difficult when the signal SIG is heavily contaminated with noise. In the case of noise RS, the correlation value is lower than, for example, in the case of a sine wave SS. However, this also depends on the amplitude of the sine wave SS. This can make it difficult to distinguish between the two categories or signal characteristics SC1 (noise signal RS) and SC3 (sine wave SS).

[0087] Overall, the correlation value c1 calculated according to formula (1) alone cannot distinguish the three signal characteristics SC1, SC2, and SC3 mentioned above, as in... Figure 5 As can be seen from it. Figure 5 In the above, a first correlation value c1, defined in formula (1), is plotted for different signal frames with different signal characteristics SC. The correlation value c1 is plotted on the x-axis, and different labels are assigned to the correlation value c1 according to the different basic signal characteristics SC.

[0088] As in Figure 5As seen in the image, the correlation values ​​c1 used for the three different signal features SC1, SC2, and SC3 overlap significantly. Therefore, it is impossible to successfully distinguish the three signal features SC1, SC2, and SC3 solely based on this calculated correlation metric or correlation value c1.

[0089] In the method according to the invention, instead of using the correlation value c1 defined in formula (1), another measure of the correlation between successive samples or sampled values ​​is used primarily, d1.

[0090] The correlation measure d1 is defined in formula (3) as follows: .

[0091] The second correlation measure d1 in formula (3) is the so-called taxicab difference between two vectors x1 and x2, where: .

[0092] exist Figure 6 In the figure, the relationship between formula (1) and formula (3) is shown for different signal characteristics or signal features SC. The signals are normalized so that the variance VAR of these signals is 1, and c1 and d1 are normalized by the quantity N.

[0093] Figure 5 The first signal shown is noise, i.e., noise signal RS, which can be generated by a random number generator. This noise signal RS satisfies the first signal characteristic SC1.

[0094] The second signal is a sinusoidal signal SS. This sinusoidal signal SS satisfies the third signal characteristic SC3.

[0095] The third signal is a mixed signal MS, which is composed of random noise RS and a sinusoidal signal SS.

[0096] To generate signals with different correlation values, randomly generated noise can be low-pass filtered at different cutoff frequencies. For example, 10 signals can be generated for each cutoff frequency, and correlation metrics c1 and d1 can be calculated for these signals. Thus, each tag represents the correlation value for a signal.

[0097] As in Figure 6 As can be seen, there is an inverse relationship between the first correlation measure c1 and the second correlation measure d1. When c1=1, d1=0. Further trends depend on the underlying signal (such as...). Figure 6(As shown in the figure). The trends of the different correlation values ​​(c1, d1) between successive sampled values ​​differ between the noise signal RS and the sinusoidal signal SS. The trend of the correlation values ​​(c1, d1) of the mixed signal MS, which is a mixture of the noise signal RS and the sinusoidal signal SS, lies between the trends of the correlation values ​​(c1, d1) for the sinusoidal signal SS and the noise signal RS.

[0098] Figure 6 As shown, the first correlation value c1 calculated according to formula (1) and the second correlation value d1 calculated according to formula (3) are both indices of the correlation between successive sampled values ​​of the signal. However, there is no fixed ratio between the first correlation value c1 calculated according to formula (1) and the second correlation value d1 calculated according to formula (3). The ratio between the first correlation value c1 and the second correlation value d1 depends on the signal characteristics SC of the signal SIG.

[0099] Therefore, calculating the first correlation value c1 according to formula (1) and the second correlation value d1 according to formula (3) allows information about the signal characteristics SC to be obtained. Depending on the ratio v of c1 to d1 (v = c1 / d1), the signal is more "noise-like" (SC1) or more "sinusoidal" (SC3). Figure 5 If the second correlation value d1 of different signals in the graph is plotted, then the distribution of the different signal features SC will be related to... Figure 5 The distributions differ. The distribution of the second correlation value d1 can be found in... Figure 7 I saw it in [the context]. Figure 7 The correlation values ​​d1 based on formula (3) are plotted not only for signal feature SC1 (noise) but also for signal feature SC3 (sine). (Not included in...) Figure 7 The signal feature SC2 (fixed signal) is plotted in the middle because the correlation value d1 used for this signal feature SC2 is always zero.

[0100] The idea behind calculating the correlation value d1 according to formula (3) is to accumulate or sum the total changes between successive samples or sampled values.

[0101] If the cumulative total change between successive sampled values ​​is large (d1 is large), then the successive sampled values ​​of the signal are only slightly correlated with each other (i.e., noise SC1).

[0102] If the cumulative total change between successive sampled values ​​is small (d1 is small), then the successive sampled values ​​are strongly correlated with each other (a more regular signal SC3).

[0103] If the cumulative total change between successive sampled values ​​is zero (d1=0), then the values ​​of successive sampled values ​​are the same, and the correlation between successive sampled values ​​is the highest (fixed signal or constant signal SC2).

[0104] In the method according to the invention, a sum of all differences is formed and compared with a reference value or threshold SW. This achieves greater robustness to outliers.

[0105] Using the method according to the invention, noise identification under normal circumstances can be achieved. The difference between successive values, calculated periodically between feedbacks, changes. Once the counter threshold of N sample values ​​for the sampling sequence is exceeded, the counter Z is reset.

[0106] Traditional algorithms also consider the duration compared to the total sample size; therefore, the difference in duration varies depending on the threshold mentioned above. However, in the method according to the invention, decisions are made periodically, i.e., after each fixed number of N samples or sampled values. The algorithm according to the invention is efficient and requires relatively little computation.

[0107] The correlation measure d1 calculated according to formula (3) can be used to identify noise in the following way. First, collect N samples or sampled values, and then calculate the correlation value d1 according to formula (3). If the calculated correlation value d1 exceeds a certain threshold SW (d1>SW), then the signal is a noise signal RS (SC1); if the calculated correlation value d1 is zero (d1=0), then the signal is fixed (SC2). In all other cases, the signal is determined to be regular (SC3).

[0108] Since collecting N sample values ​​requires storage space, it is also possible to calculate the second correlation value d1 iteratively. The iterative calculation of the second correlation value d1 is given in formula (5): .

[0109] In the iterative calculation given in formula (5), it is only necessary to store the value d1,n-1 of the previous length and the last sample xn-1. Using these two values ​​and the current value xn, the second correlation value d1,n can be calculated.

[0110] In one embodiment of the method according to the invention, the correlation value d1 is calculated according to formula (3), wherein all samples or sampled values ​​are first collected and stored in a buffer memory before the correlation value d1 is calculated. However, since this requires storage capacity and occupies a certain amount of storage space, in a possible alternative embodiment of the method according to the invention, the correlation value d1 is calculated iteratively according to formula (5).

[0111] Although the present invention has been fully described above based on preferred embodiments, the present invention is not limited thereto, but can be modified in various ways.

Claims

1. A method for identifying a signal characteristic of a sensor signal from a sensor (2), the method comprising the following steps: sampling (SI) the sensor signal (SIG) at a sampling rate to produce sample values (x); calculating (S2) a correlation measure (dl) for the sensor signal (SIG) by summing absolute differences between two sample values (xi+1; xi) of a sample sequence of the sampled sensor signal, which sample sequence comprises a determined number N of sample values, successive to each other; and determining (S3) a signal characteristic (SC) of the sensor signal (SIG) on the basis of the calculated correlation measure (dl). comparing the calculated correlation measure (dl) with a threshold value (SW). classifying the sensor signal (SIG) as noise-like if the calculated correlation measure (dl) is higher than the threshold value (SW). classifying the sensor signal (SIG) as regular if the calculated correlation measure (dl) is lower than the threshold value (SW).

2. The method for identifying a signal feature of a sensor signal from a sensor according to claim 1, wherein classifying the sensor signal (SIG) as constant if the calculated correlation measure (dl) is zero.

3. The method for identifying a signal feature of a sensor signal from a sensor according to claim 2, wherein temporarily storing sample values of the sample sequence and carrying out the calculation (S2) of the correlation measure (dl) on the basis of the temporarily stored sample values of the sample sequence.

4. The method for identifying a signal feature of a sensor signal from a sensor according to claim 3, wherein carrying out the calculation (S2) of the correlation measure (dl) in an iterative manner without temporarily storing sample values of the sample sequence.

5. The method for identifying a signal feature of a sensor signal from a sensor according to claim 4, wherein, calculating a further correlation measure (cl) for the sensor signal by summing products of two sample values (xi+1; xi) of a sample sequence of the sampled sensor signal, which sample sequence comprises a determined number N of sample values, successive to each other.

6. The method for identifying a signal feature of a sensor signal from a sensor according to any one of claims 1 to 5, wherein storing the two calculated correlation measures (dl, cl) for the sensor signal (SIG) in pairs.

7. The method for identifying a signal feature of a sensor signal from a sensor according to any one of claims 1 to 5, wherein calculating a ratio between the two correlation measures (dl, cl).

8. The method for identifying a signal feature of a sensor signal from a sensor according to any one of claims 1 to 7, wherein further specifying a signal characteristic (SC) of the sensor signal (SIG) in dependence on the calculated ratio between the two correlation measures (dl, cl) for a determined value of one of the two correlation measures.

9. The method for identifying a signal feature of a sensor signal from a sensor according to claim 8, wherein, evaluating on the basis of the ascertained signal characteristic (SC) of the sensor signal (SIG) whether the sensor (2) is working fault-free or has a fault.

10. The method for identifying a signal feature of a sensor signal from a sensor according to claim 9, wherein, automatically activating an operating mode for handling the identified fault of the sensor (2) upon identifying a fault of the sensor (2).

11. The method for identifying a signal feature of a sensor signal from a sensor according to claim 10, wherein the sensor signal (SIG) is output by a MEMS sensor of a gyroscope.

12. The method for identifying a signal feature of a sensor signal from a sensor according to any one of the preceding claims 1 to 11, wherein, 15. An apparatus (1) for identifying a signal characteristic (SC) of a sensor signal from a sensor (2), the apparatus comprising: a sampling unit (1A) for sampling the sensor signal at a sampling rate to produce sample values (x); a calculation unit (1B) for calculating a correlation measure (dl) for the sensor signal (SIG) by summing absolute differences between two sample values (xi+1; xi) of a sample sequence of the sampled sensor signal, which sample sequence comprises a determined number N of sample values, successive to each other; and a determination unit (1C) for determining a signal characteristic (SC) of the sensor signal (SIG) on the basis of the calculated correlation measure (dl).

13. The method for identifying a signal feature of a sensor signal from a sensor according to claim 12, wherein, ​ 14. The method for identifying a signal feature of a sensor signal from a sensor according to any one of the preceding claims 1 to 13, wherein, ​ ​ ​ a computing unit (1 B) which calculates a correlation measure (d1) for the sensor signal by summing up the absolute difference between two successive sample values (xi+1; xi) of a sample sequence of the sampled sensor signal, wherein the sample sequence comprises a determined number N of sample values; and a determining unit (1 C) which determines a signal characteristic (SC) of the sensor signal on the basis of the calculated correlation measure (d1).

16. The device for identifying a signal feature of a sensor signal from a sensor according to claim 15, wherein The determining unit (1 C) has a comparator which compares the correlation measure (d1) calculated by the computing unit (1 B) with a threshold value (SW), wherein the sensor signal is classified as noise-like by the determining unit (1 C) if the calculated correlation measure (d1) is higher than the threshold value (SW), wherein the sensor signal is classified as regular by the determining unit (1 C) if the calculated correlation measure (d1) is lower than the threshold value (SW), and wherein the sensor signal is classified as constant by the determining unit (1 C) if the calculated correlation measure (d1) is zero.

17. The device (1) for identifying a signal characteristic (SC) of a sensor signal from a sensor (2) according to claim 15 or 16, wherein The apparatus (1) has a buffer memory for temporarily storing the sample values of the sample sequence generated by the sampling unit (1 A).

18. The device (1) for identifying a signal characteristic (SC) of a sensor signal from a sensor (2) according to any one of claims 15 to 17, wherein The computing unit (1 B) of the apparatus (1) calculates a further correlation measure (c1) for the sensor signal by summing up the product of two successive sample values (xi+1; xi) of a sample sequence of the sampled sensor signal, wherein the sample sequence comprises a determined number N of sample values.

19. The device (1) for identifying a signal characteristic (SC) of a sensor signal from a sensor (2) according to claim 18, wherein The two correlation measures (d1, c1) calculated by the computing unit (1 B) for the sensor signal are stored in pairs in a data memory of the apparatus (1).

20. The device (1) for identifying a signal characteristic (SC) of a sensor signal from a sensor (2) according to claim 18 or 19, wherein The computing unit (1 B) of the apparatus (1) calculates a ratio between the two correlation measures (d1, c1).

21. The device (1) for identifying a signal characteristic (SC) of a sensor signal from a sensor (2) according to claim 19 or 20, wherein The determining unit (1 C) of the apparatus (1) further specifies the signal characteristic (SC) of the sensor signal on the basis of the ratio between the two correlation measures (d1, c1) calculated by the computing unit (1 B) for the determined value of one of the two correlation measures.

22. The device (1) for identifying a signal feature (SC) of a sensor signal from a sensor (2) according to any one of claims 15 to 21, wherein The evaluation unit of the apparatus (1) evaluates on the basis of the determined signal characteristic (SC) of the sensor signal whether the sensor is functioning faultlessly or has a fault.

23. The device (1) for identifying a signal characteristic (SC) of a sensor signal from a sensor (2) according to claim 22, wherein The evaluation unit outputs an activation signal for activating an operating mode for handling the fault of the sensor if a fault of the sensor is identified on the basis of the determined signal characteristic (SC).

24. A system (SYS) comprising at least one sensor (2) and an apparatus (1) for identifying a signal characteristic (SC) of a sensor signal from the sensor (2) according to any one of claims 15 to 23.

25. The system (SYS) according to claim 24, wherein, The sensor (2) is a MEMS sensor.

Images