Accelerometer with an automatic calibration
The invention relates to a device for measuring an activity of a person comprising a multi-axes sensor, comprising accelerometers arranged corresponding to orthogonal axes, said multi-axes sensor being arranged to generate electrical signals as a function of the person's motion.
The invention further relates to a method for calibrating a multi-axes sensor comprising accelerometers arranged corresponding to orthogonal axes.
The invention still further relates to a system for monitoring a physiological condition of a patient.
A device as set forth in the opening paragraph is known from WO0247465. The known multi-axes sensor is arranged to monitor a physical activity of an individual, when being worn by the individual. Under working conditions the multi-axes sensor produces a signal characteristic to an inclination of the person with respect to the earth's gravitational field. In order to determine the acceleration of the individual correctly the signal of the sensor must be corrected for possible inaccuracies in the calibration of the sensor, namely a drift if the absolute value for a gain and an offset of each axis must be periodically corrected for. In the accelerometers of the class known from WO0247465 said correction, further referred to as calibration is performed in a special mode, known as calibration mode. In the calibration mode the sensor signal has to be analyzed for a known alignment of the sensor's sensitive axes with the direction of the earth's gravitational field.
A disadvantage of the known device is the necessity for a separate calibration routine, where the device and/or the individual wearing the device must assume a predetermined orientation in space leading to a cumbersome calibration procedure resulting in additional costs and efforts during the manufacturing on one hand, and in a loss of time due to the calibration during exploitation on the other hand.
It is an object of the invention to provide the device as set forth in the opening paragraph where the calibration procedure is simplified and can be enabled without losing time.
The device according to the invention is characterized in that said multi-axes sensor comprises calibration means arranged to identify suitable periods for calibration of a gain and an offset of said accelerometers and to perform said calibration in an automatic mode. The proposed calibration is based on the insight that for the rest periods of the individual the norm of the net acceleration vector as measured by the device equals to lg, independent of the orientation of the sensitive axes of the accelerometers. Therefore, the signals from the corrected accelerometers, when set out in the orthogonal XYZ-space would form a sphere. On the contrary, the signals from the non-corrected accelerometers would form an ellipsoid. The computing means of the device according to the invention are arranged to detect the rest periods, characterized by minimal signal changes over a period of, for example, a few seconds. When such a rest period is detected, the calibration means initiates a calibration routine. The calibration routine, thus, can be carried out automatically by analyzing the accelerometer output signals and by fitting the shape of the elipsoid to the sphere.
An embodiment of the device according to the invention is characterized in that the calibration means comprise computing means arranged to carry out said calibration according to a predetermined computing algorithm. It is understood that the characteristics of the axes can be approximated by the equations: x = kx - ax + xQ y = ky - ay +y0
x, y and z denote the measured accelerometer outputs, whereas ax, ay and az are the real accelerations, xo, yo and zo are the offsets and kx, ky and kz the gain factors for each accelerometer.
Given that the three accelerometers which have axes which are perpendicular to each other the norm of the acceleration vectors at rest equals the earth gravitation:
Inserting the above equations gives:
The equation states that all non-corrected measurement triples (X
R, y
R, Z
R) acquired during rest are located on the surface of an ellipsoid. From that the basic calibration parameters can be deduced, for example by means of a numerical fit of an ellipsoid to the measured points. An example of a suitable numerical fit algorithm is a least square fit method. The resulting six ellipsoid parameters are the desired calibration values for the gain and offset values for each axis. It must be noted, that for the ellipsoid to be fitted stable and reliably to the measured triples they must be well-spread over the XYZ-space. This is obtained in the device according to the invention by monitoring the output signals by the computing means and by selecting those signals corresponding to the rest periods, which are substantially well-spread in the XYZ space. By providing the device with computing means arranged to perform the analysis of the accelerometer running signal and to modify the gains and the offset of the multi-axes sensor when the signals do not comply with the shape of the sphere, an accurate automatic real-time calibration is enabled. The device according to the invention is particularly suited to be used in a system for monitoring of a physical condition of the individual. During the monitoring application the accelerometer signals are collected and are used to add to the information taken from other body sensors. In case a sensor tuned to a specific physiological condition reports an alarm situation, the alarm can be confirmed by the signals from the accelerometer. In this way a reliable monitoring system is achieved with a low rate of false alarms. A further embodiment of the device according to the invention is characterized in that the calibration means further comprises a trigger unit arranged to trigger the calibration based on a preset criterion. An automatic calibration and further recalibration under operating conditions can be triggered by the trigger unit of the calibration means in case the resting accelerations deviate significantly from lg. Preferably, the calibration means are arranged to comply with a threshold uncertainty, upon excess of which the automatic calibration is enabled. Preferably, the threshold uncertainty is stored in a look-up table of a device's microprocessor and serves as the preset criterion.
A method for calibrating a multi-axes accelerometer according to the invention is characterized in that said method comprises the steps of: - determining static periods in the accelerometer signals, said periods representing significantly different orientations of the accelerometer in the XYZ-space; assigning measurement points in the XYZ-space corresponding to thus obtained accelerometer signals;
- performing a numerical fit of an ellipsoid to the measurement points;
- assigning the calibration values to the ellipsoid parameters.
According to the method according to the invention the sensor is supplied with a reliable calibration routine arranged to carry out the accelerometer self-calibration in an automatic mode. In order to enable said calibration the sensor is preferably supplied with calibration means comprising computing means arranged to carry out the steps of the above method.
These and other aspects of the invention will be explained with reference to the Fig. Fig. 1 shows a schematic view of an embodiment of the device according to the invention.
Fig. 2 shows a schematic view of an embodiment of a method according to the invention. Fig. 3 shows a schematic view of an embodiment of a system for monitoring a physiological condition of an individual according to the invention.
Fig. 1 shows a schematic view of an embodiment of the device according to the invention. The device 1 is intended to be worn by the individual I, or to be implanted into the individual I. The device 1 comprises a multi-axes sensor 1 with a number of DC accelerometers, arranged in parallel to a corresponding X-, Y- or Z-axes of the Cartesian coordinate system. Preferably, the multi-axes sensor 1 comprises three DC accelerometers, schematically represented by 3X, 3Y, 3Z in Fig. 1, however in less preferable configuration the multi-axes sensor 1 can be arranged with two DC accelerometers. The multi-axes sensor 3 is arranged to determine an orientation of the individual's virtual axes XI, Yl, Zl I with respect to the earth's gravitational field g. This determination is carried out by the microprocessor 9. The microprocessor 9 is arranged in the device 1 to process the signals from the accelerometers 3X, 3Y and 3Z. The method of determination of the alignment is known per se from US6044297 and will not be explained in detail. The device 1 further comprises calibration means 5 arranged to correct the signals from the multi-axes sensor for possible calibration drift in the values for gain and offset of the accelerometers. In order to perform the calibration in an automatic mode, the device 1 according to the invention comprises computing means 7, which are arranged to initiate the calibration in certain events.
An example of such an event is the drift of the gain and offset value beyond a preset accuracy threshold. The value of the threshold can be stored in a look-up table 11 and can be addressed by the computing means 7 during a real-time operation. The computing means according to the invention is arranged to analyze the signals from the accelerometers 3X, 3 Y, 3Z and to detect a rest condition. The rest condition of the individual is characterized by a small signal change, for example when the absolute value of the signal is comparable to noise. When the rest condition is determined, the computing unit processes the signals from the multi-axes sensor and sets them out in the three-dimensional space. In situations, when the data do not comply with a sphere, the trigger unit 13 triggers the automatic calibration of the sensitive axes of the device 1. The calibration is carried out by the computing means 7 according to a suitable algorithm. An example of such suitable algorithm is a numerical fit to the signals, corresponding to rest conditions, fitting them to an ellipsoid. The parameters of the ellipsoid are further used as the calibration parameters by the microprocessor 9. Preferably, the calibration parameters are written in a look-up table 11 ' and are used until a following calibration is initiated. The steps of the calibration routine will be explained with reference to Fig. 2.
Fig. 2 shows a schematic view of an embodiment of a method according to the invention. The multi-axes sensor 3 provides raw signals (x,y,z) as output from the three accelerometers 3X, 3 Y, 3Z. The raw signals (x,y,z) are processed by the computing means and are compared to a criterion 6 of a rest condition by the computing means at step 7. If the rest condition is satisfied, the raw signals get a flag corresponding to the rest condition (XR, VR, ZR) and are made available for the correction step 8. At this step the current calibration parameters, stored in the look-up table 11' are addressed and actual accelerations (ax, ay, az) according to a suitable algorithm are computed. The computed accelerations are checked against a validity criterion step 12. A suitable validity criterion is a comparison of the norm of computed acceleration with the earth's gravitational field g. In case the norm of the accelerations equals to g with an acceptable uncertainty, stored in a look-up table [not shown], the calibration parameters are accepted and the system proceeds to a step 15 of pending for signal analysis corresponding to a physical activity of the individual. In case the current calibration parameters result in computed accelerations exceeding the acceptable inaccuracy, the system moves to step 17 of recalibration, where the rest data are fitted to an ellipsoid, and the new calibration parameters are written into the look-up table 11'. These new calibration parameters are further addressed by the computing means to calculate the net acceleration in case the individual resumes physical activity.
Fig. 3 shows a schematic view of an embodiment of a system for monitoring a physiological condition of an individual comprising the device according to the invention. Here an embodiment of the components of the user-side of the system are presented. The user-side 20 comprises monitoring means 22 arranged to monitor a physiological condition of the user. The monitoring means 22 comprise a set of electrodes 21 arranged on the body of the user to pick-up a signal characteristic of the physiological signal, for example an ECG signal, a body temperature, respiration rate, encephalogram, etc. Additionally, the monitoring means 22 can comprise a sensor 21' arranged to monitor a signal not directly related with a targeted physiological condition. An example of such a sensor is a multi-axes accelerometer. The monitoring means 22 are arranged to perform a continuous monitoring of a physiological condition of the user and are further arranged to provide a corresponding signal to a front-end electronics 27 of the user-side 20 of the monitoring system. The monitoring means 22 and the front-end electronics 27 are worn on the body of the user. The front-end electronics 27 is arranged to analyze signals from the sensors in order to derive a feature characteristic to an abnormality in the physiological condition of the user. For that purpose the front-end electronics 27 comprise a preamplifier and analogue processing circuit 23, an ADC unit 24, a μ-processor 25, detection means 28, alarm means 29 and transmission means 31. The detection means 28 comprise a sensor signal interpretation unit 32 and feature extraction means 33. The user-side 20 of the system operates as follows: the monitoring means 22 acquire the raw data from sensors 21 and 21 ', which are delivered to the front-end electronics 27. The front-end electronics provides means for receiving the signals from the monitoring means, performs suited analogue processing by means of the analogue processing circuit 23. The processed raw data is converted into a digital format by means of the ADC 24 and is forwarded by the μ-processor 25 to the detection means 28, where the condition of the user is being analyzed. For example, for cardiac applications the detection means 28 can comprise a per-se known QRS-detector to determine R-R peak intervals in heart cycles. The detection means 28 can be arranged to comprise a sensor signal interpretation unit 32 arranged to derive a feature in the signal characteristic to an abnormal physiological condition of the user. For example, for cardiac or cranial applications said feature can be a frequency of the signal, for heamodynamic studies said feature can be a threshold value of a blood pressure and so on. It is also possible that more than one feature is assigned per monitored physiological condition. In this case the features can be ranked up according to the severity of the abnormality of the physiological condition being monitored. For example, for cardiac applications, a minor change in the cardiac cycle can be recognized as an alarm of the lowest
category, whereas an occurrence of arrythmia or fibrillation can be ranked higher. The alarm signal can be ranked according to the rank of the feature. In both situations, the value of the feature or the features can be stored in a look-up table (not shown) of the memory unit 34. Additionally, the system can be arranged as a self-learning system, where the threshold value for the feature is being adjusted and stored in the look-up table in cases a pre-stored value does not correspond to an abnormal condition for a particular user.
In case the detection means 28 detects the abnormal condition, and a signal from the multi-axes accelerometer reports a rest condition of the user (corresponding to fainting, for example) a signal is sent to the alarm means 29 to generate an alarm, which is transmitted by the transmitting means 31 , for example by means of a RF-link. The alarm signal is transmitted to a home station in case the user experiences an abnormality at home, or, alternatively to a mobile station for locations of the user away from home. From the respective station the emergency center is informed and is provided with the exact position of the customer (at home / actual position outside home). The alarm center takes over the management of the emergency and informs the respective communal or medical sites about the emergency, the location, patient data and the probable diagnose. This enables minimization of a response time for the medical assistance and gives an improved chance to save customers' lives.
It must be noted that the system as described with reference to Fig. 3 can be suited to monitor a physiological condition of a user under normal, not related to any disease conditions, for example for fitness purposes. In this case the alarm generation means is arranged to alarm the user if his heartbeat, temperature, or respiration rate, accordingly exceeds a healthy level. The multi-axes accelerometer 21 ' is utilized in a fitness-related application to provide additional data upon the physical activity of the user. This application is known per se and will not be explained in detail. The advantage of using a multi-axes sensor according to the invention is due to the fact this device ensures accurate durable monitoring of a physiological conditions without a necessity to interrupt the monitoring for purposes of the sensor calibration. By triggering a self-calibration during rest periods of the user the multi-axes accelerometer calibration data are being updated adding to a reliability of the system as a whole.