KR20100078754A - Apparatus and method for detecting falldown - Google Patents

Abstract

PURPOSE: A device and a method for sensing a fall are provided to prevent an accident by sensing a fall by alarming whether a fall is generated or not to outside. CONSTITUTION: A device for sensing a fall comprises a sensing unit(10), a signal processor(20), a fall determining unit(30), and a warning unit(40). The sensing unit senses a vibration signal through sensors attached on a floor. The signal processor converts the sensed vibration signal into a frequency area and extracts the feature of the converted vibration signal. The fall determining unit determines whether the fall is generated or not. The warning unit generates an alarming signal depending on the determined result.

Description

Fall detection device and fall detection method {Apparatus and method for detecting falldown}

One aspect of the present invention relates to a sensing technique, and more particularly to a fall sensing technique.

In recent years, as the aging population is increasing, elderly people or people with discomfort are increasing, and unfortunate accidents due to falls have occurred to these people. In particular, the elderly often live alone. Older people, unlike young people, are at risk from minor injuries. Indeed, weak elderly people are difficult to balance or fall off the sense of balance, in this case, depending on a certain object can fall, secondary damage can lead to a big accident.

In addition, young people are often bruised if they fall forward, but the elderly may be at risk of brain bleeding, which can be dangerous. Falling brain damage can result in long-term absence and affect both thinking and human function.

In particular, the Korea Safety and Health Association surveyed 357 elderly people aged 65 or older in Seoul on September 27-28, 2007, and found that 8 out of 10 elderly people had a fall accident. Therefore, it is urgent to introduce fall detection technology to prepare for a fall accident.

According to an aspect, a fall detection device and a fall detection method for preventing a fall accident are proposed.

According to an aspect of the present invention, a fall detection apparatus includes a sensing unit configured to detect a vibration signal through sensors attached to a floor, converting a detected vibration signal from a time domain into a frequency domain, and extracting a feature on a vibration signal converted into a frequency domain The signal processing unit includes a fall determination unit that determines whether or not a fall on the floor by comparing an input pattern corresponding to the extracted feature with a pre-stored standard pattern, and an alarm unit that generates an alarm signal according to the determination result.

Meanwhile, a fall detection method according to another aspect includes detecting a vibration signal through sensors attached to a floor, converting the detected vibration signal from a time domain into a frequency domain, and extracting a feature on the vibration signal converted into the frequency domain. Comprising the step of comparing the input pattern corresponding to the extracted feature with a pre-stored standard pattern to determine whether the fall on the floor and generating a warning signal according to the determination result.

As described above, according to an embodiment of the present invention, the fall accident can be prevented through the fall detection. In other words, in order to prevent the fall accidents that frequently occur to the elderly, it is possible to actively prepare for the second accident caused by the fall by accurately determining whether the fall and alarming the fall. Furthermore, by detecting falls and monitoring them in real time, it is possible for the elderly to live their daily lives with peace of mind and to reassure their families.

Hereinafter, with reference to the accompanying drawings will be described embodiments of the present invention; In the following description of the present invention, if it is determined that detailed descriptions of related well-known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. In addition, terms to be described below are terms defined in consideration of functions in the present invention, which may vary according to intention or custom of a user or an operator. Therefore, the definition should be made based on the contents throughout the specification.

1 is a block diagram of a fall detection device 1a according to an embodiment of the present invention. Referring to FIG. 1, the fall detection apparatus 1a according to an embodiment includes a sensing unit 10, a signal processing unit 20, a fall determination unit 30, an alarm unit 40, and a storage unit 50. .

The sensing unit 10 detects a vibration signal through sensors attached to the floor. The sensor can be any type that can detect vibration. In this case, the sensor may be attached to a floor of an indoor room, a bedroom, a living room, a bathroom or a toilet, a staircase in the house, but is not limited thereto. The sensor may be exposed or buried exposed to the outside of the floor. In particular, if the sensor is buried, it may be a built-in (built-in) form is installed from the beginning when the floor construction. The vibration signal detected by the sensing unit 10 may be transmitted to the signal processing unit 20 to be described later through a communication method such as RS485, but the communication method is not limited thereto.

The signal processor 20 converts the vibration signal detected by the sensing unit 10 from the time domain to the frequency domain and extracts features from the vibration signal converted into the frequency domain. do. In this case, the vibration signal may be transformed from the time domain to the frequency domain using a fast Fourier transform (FFT) or a Discrete Fourier transform (DFT).

The signal processor 20 extracts a feature of the vibration signal converted into the frequency domain, and the feature may be a distinguishable aspect, quality, or characteristic of the vibration signal inherent in the vibration signal. The feature may be a symbolic symbol or a numerical value. When the feature has one or more numerical values, the feature may be represented by a d-dimensional column vector called a feature vector. The d-dimensional space in which the feature vector is defined is called a feature space. Therefore, the objects to be recognized are represented by the points formed by the feature vector in the feature space. In particular, a picture of features represented by points represented in a feature space is called a "scatter plot." Furthermore, the signal processor 20 may classify the extracted features. According to an exemplary embodiment, the signal processing unit 20 may classify the features using a nonlinear classifier. Nonlinear classifiers may use supervised learning-based methods, such as k nearest neighbor and neural networks.

On the other hand, the fall determination unit 30 determines whether the fall on the floor by comparing the input pattern corresponding to the features processed and extracted by the signal processing unit 20 with the standard pattern stored in the storage unit 50 to be described later do. Here, the pattern corresponds to a collection of features of the detected signal. Features and patterns are similar concepts, but features can be seen to form a pattern. In pattern recognition, a pattern may be represented by a pair of variables {x, ω} consisting of the observed feature vector x and the class ω to which the feature vector belongs. In this case, a class is also called a category, a group, or a label, and refers to a pattern set that shares a common set of attributes or features generated from the same source.

The fall determination unit 30 performs pattern matching for comparing the standard pattern stored in the storage unit 50 with the input pattern. In this case, when the input pattern corresponding to the extracted feature of the vibration signal is similar to the standard pattern, it may be determined that the user who has detected the vibration on the floor has fallen. The user may include everyone, but may specifically target the elderly and the like.

The storage unit 50 stores a standard pattern for vibration. In this case, the stored standard pattern may be a pattern set differently according to body characteristics of each user and media characteristics of the floor. The user's body characteristics may include at least one of a user's weight, body fat, height, age or gender. This is because the vibration signal detected at the fall according to the user's body characteristics is different, the storage unit 50 stores the frequency correlation according to the user's body characteristics. In addition, the standard pattern sets a pattern of a person and an object differently, and allows the fall determination unit 30 to determine whether the vibration signal is generated by a person or an object. Accordingly, it is possible to reduce the error of the fall detection of the fall detection device 1a and to improve the accuracy.

The standard pattern stored in the storage unit 50 may be directly input by the user, and may be updated from time to time according to a change in the user's environment. Alternatively, the standard pattern may be set using standard data of ordinary people corresponding to body standards. For example, the standard pattern may be set according to the criteria such as dry type, standard type, obesity type, etc. corresponding to the body standard. On the other hand, the standard of the floor can be set differently according to the medium that can form the floor, such as wood, concrete, stone. This is because the vibration signal for determining the fall may vary depending on the medium of the floor. In addition, the storage unit 50 may include an installation map of the sensors to determine the fall position.

On the other hand, the alarm unit 40 generates an alarm signal in accordance with the determination result of the fall determination unit 30. That is, when the determination result of the fall determination unit 30 determines that the user has fallen, an alarm signal may be generated to the outside. The alarm signal generation of the alarm unit 40 may be in the form of CDMA transmission, but is not limited thereto and may include both current and future communication means.

2 is a block diagram of the fall detection device 1b according to another embodiment of the present invention. 2, the fall detection apparatus 1b according to another embodiment includes a sensing unit 10, a signal processing unit 20, a fall determination unit 30, an alarm unit 40, and a storage unit 50. The apparatus may further include at least one of a preliminary determination unit 60, a pre-signal processing unit 22, a post determination unit 70, an emergency signal transmission unit 80, and a monitoring unit 90. Detailed descriptions of the sensing unit 10, the signal processing unit 20, the fall determination unit 30, the alarm unit 40, and the storage unit 50 are omitted in FIG. 1.

The emergency signal transmitter 80 generates an emergency signal for notifying a danger when a response signal is not input within a preset time period for the alarm signal generated through the alarm unit 40. In this case, the response signal within a preset period may be input through a wired or wireless type of recocon. The remote controller may be implemented in a form that can be carried by a user using a communication means such as ZigBee in the wireless form, and may be attached to a wall of a room in a wired form, but the implementation form and the communication means are not limited thereto. In addition, the emergency signal generated through the emergency signal transmitter 80 may be a signal that may be connected to an emergency treatment facility, such as a designated hospital, or a signal that may be connected to a user's guardian.

The preliminary determination unit 60 determines whether the vibration signal detected by the sensing unit 10 is greater than or equal to a predetermined threshold value, and when the vibration signal is greater than or equal to the threshold value, the signal processing unit 20. To send. That is, the vibration signal detected by the sensing unit 10 is not directly transmitted to the signal processing unit 20, but only to the signal processing unit 20 when the threshold value is greater than or equal to the threshold value. This saves power and speeds up the fall detection process.

On the other hand, the post-determination unit 70 is located between the fall determination unit 30 and the alarm unit 40, the fall or fall based on the detection of the vibration signal during the predetermined period after the fall-off determination by the fall determination unit 30 Judge again. At this time, if the fall determination unit 30 is a means for primarily determining the fall of the user, the post-determination unit 70 may be referred to as a second determination means for judging the fall of the user again. The reason for detecting the vibration signal for a predetermined period after the fall determination is determined by the fall determination unit 30 is to determine whether the user finally falls by determining whether activity (movement) is detected after the fall. Here, if the vibration signal is not detected for a predetermined period, it is finally determined as a fall and controls to generate an alarm signal through the alarm unit 40.

The signal preprocessor 22 removes noise with respect to the vibration signal detected by the sensing unit 10. Meanwhile, the monitoring unit 90 monitors the input pattern processed through the signal processing unit 20 to inform the monitored result to the outside. In this case, the guardian of the user can check the user's daily life through the monitoring program. Accordingly, it is possible to monitor in real time the user's abnormality other than the fall.

3A to 3D are reference diagrams for describing a signal processing process of the signal preprocessor of FIG. 2 according to an embodiment of the present invention.

The sensor that detects the vibration signal is affected by noise, and the noise may appear as shown in FIG. 3A. If the noise affects the vibration, the pattern of the vibration itself may change under the influence of the noise. This change is disadvantageous when classifying vibration patterns. Referring to FIG. 3B, the noise generally appears in the form of low frequency and the frequency band is mostly distributed below 6 Hz.

According to an embodiment of the present invention, a high-pass filter or a band-pass filter is used to remove the influence of the noise described above. A high-pass filter is a filter that passes only frequencies above the specified frequency by attenuating the signal on the low frequency side. In addition, a band-pass filter is a filter that passes signals existing in a specific range of frequencies and removes signals outside this range.

FIG. 3C shows a high pass filter with a cut-off frequency of 6 Hz according to one embodiment, and FIG. 3D shows a result after using the high pass filter according to an embodiment for the signal of FIG. 3B. Referring to FIG. 3D, it can be seen that noise having a large influence on the amplitude can be removed by using the high pass filter.

4A and 4B are reference diagrams for explaining a frequency domain conversion process of a vibration signal of the signal processor of FIGS. 1 and 2 according to an exemplary embodiment of the present invention.

4A shows the mannequin falling data obtained when the falling experiment was performed in different time zones. Referring to FIG. 4A, the patterns of the two data are similar but there is a difference in the position of the pattern in the time domain. To match two patterns in the time domain, the two signals must be correctly mapped to occur in the same time zone, so that matching is possible. However, mapping two signals accurately in the time domain to occur in the same time zone is more difficult than pattern recognition.

When the signals in the time domain are converted into the frequency domain, signals having the same characteristics are located at the same position in the frequency domain, thereby solving the mapping problem of signals acquired at different time zones. According to an embodiment, a signal of a time domain may be converted into a frequency domain using a fast Fourier transform (FFT). 4B shows the result when the signal shown in FIG. 4A is FFTed. Referring to FIG. 4B, it can be seen that similar patterns are concentrated in the same frequency domain.

5 is a reference diagram for explaining a feature extraction process of the signal processor of FIGS. 1 and 2 according to an exemplary embodiment of the present invention.

After converting the vibration signal into the frequency domain, the data in the frequency domain is used as a feature in pattern recognition. The higher the dimension of the feature space, the more similar information there is and the higher the correlation between the information, making it difficult to accurately identify it. In addition, the increase in dimensions causes problems of dimensionality (curse of dimensionality) and small sample size (SSS). For this reason, pattern reduction and feature extraction are essential for more accurate identification. Referring to FIG. 4B, it can be seen that data in the frequency domain is usefully distributed between 0 and 50 Hz in distinguishing patterns. According to an embodiment, the first 256 data are extracted from the frequency domain data after a 4096-point FFT on the falling data, and used for pattern recognition (frequency range: 0 to 62.5 Hz * maximum frequency range with useful information *). ). 5 shows 256 data extracted as a feature. However, the above-described embodiment is only one embodiment to help understanding of the present invention, and various other embodiments are possible.

Meanwhile, the features extracted in the feature extraction process may be classified through a classifier. For example, the features extracted in the feature extraction process have 256 data, which are represented as a point in the 256-dimensional feature space. When there are other kinds of features similar to the features to be classified in the feature space, the decision regions are generally nonlinear, so classification with a linear classifier causes many misperceptions. According to an embodiment of the present invention, a pattern may be classified using a k nearest neighbor (KNN) classifier, which is a kind of nonlinear classifier. The KNN classifier, which is the minimum distance identification method, selects the k standard patterns closest to the input pattern and classifies the patterns by applying the majority rule. The KNN classifier is simple in structure and shows high classification performance even in the pattern classification in which the crystal regions of standard patterns overlap. In addition, according to an embodiment, the features extracted in the feature extraction process may be classified through a supervised learning based method including a neural network. The neural network is also called an artificial neural network, and according to an embodiment, may be applied to pattern recognition in a situation where linear separation is impossible.

6 is a flowchart illustrating a fall detection method according to an embodiment of the present invention.

Referring to FIG. 6, a vibration signal is detected through a sensor attached to the floor (S100). Subsequently, the detected vibration signal is signal processed, and the vibration signal is converted from the time domain to the frequency domain, and features are extracted from the vibration signal converted into the frequency domain (S110). Subsequently, a fall on the floor is determined by comparing an input pattern corresponding to the extracted feature with a pre-stored standard pattern (S120). Then, the alarm signal is generated according to the determination result (S130).

Furthermore, the method may further include generating an emergency signal for notifying a danger when a response signal is not input within a preset period with respect to the alarm signal generated in the alarm signal generation step S130. In this case, the response signal within a preset period may be input through a wired or wireless remote controller. Furthermore, before the signal processing step S110, the method may further include determining whether the detected vibration signal is greater than or equal to a preset threshold value and processing the signal when the detected vibration signal is greater than or equal to the threshold value.

Furthermore, the method may further include determining whether the fall has occurred, based on whether the vibration signal is detected during a predetermined period after the fall determination, between the fall determination step (S120) and the alarm signal generation step (S130).

Further, the input pattern signal-processed through the signal processing step (S110), the determination result by the fall determination step (S120), the alarm signal generated in the alarm signal generation step (S130), the pre-stored user characteristics, the fall history, the fall It may further include a monitoring step for notifying the occurrence point.

The embodiments of the present invention have been described above. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

1 is a block diagram of a fall detection device according to an embodiment of the present invention,

2 is a block diagram of a fall detection apparatus according to another embodiment of the present invention,

3A to 3D are reference views for explaining a signal processing process of a signal preprocessing unit according to an embodiment of the present invention;

4A and 4B are reference views for explaining a frequency domain conversion process of a vibration signal of a signal processing unit according to an embodiment of the present invention;

5 is a reference diagram for explaining a feature extraction process of a signal processor according to an embodiment of the present invention;

6 is a flowchart illustrating a fall detection method according to an embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

1a, 1b: Fall detection device 10: Sensing unit

20: signal processor 22: signal preprocessor

30: fall judgment unit 40: alarm unit

50: storage unit 60: preliminary determination unit

70: post decision unit 80: emergency signal transmission unit

90: monitoring unit

Claims (15)

Sensing unit for detecting the vibration signal through the sensors attached to the floor; A signal processor converting the sensed vibration signal from the time domain into the frequency domain and extracting a feature on the vibration signal converted into the frequency domain; A fall determination unit that determines whether a fall occurs on the floor by comparing an input pattern corresponding to the extracted feature with a pre-stored standard pattern; And Fall detection device comprising an alarm for generating an alarm signal in accordance with the determination result. The method of claim 1, And an emergency signal transmitter configured to generate an emergency signal informing of danger if a response signal is not input within a preset period of time with respect to the alarm signal generated through the alarm unit. The method of claim 2, And the response signal within the preset period is input through a wired or wireless type remote controller. The method of claim 1, The pre-stored standard pattern is a pattern set differently according to the user's body characteristics and the floor medium characteristics. The method of claim 4, wherein The user's body characteristics include at least one of the user's weight, body fat, height, age, gender. The method of claim 1, And a preliminary determination unit which determines whether the detected vibration signal is greater than or equal to a predetermined threshold value and outputs the detected vibration signal to the signal processing unit when the detected vibration signal is greater than or equal to the threshold value. According to claim 1, Between the fall determination unit and the alarm unit, And a post-determination unit for re-determining whether or not the fall is based on whether the vibration signal is detected during a preset period after determining whether the fall is determined by the fall determination unit. The method of claim 1, The signal processor classifies the extracted feature through a nonlinear classifier, The fall detection unit, characterized in that the falling fall on the floor by comparing the input pattern corresponding to the features classified by the non-linear classifier with a pre-stored standard pattern. The method of claim 8, And the nonlinear classifier uses a supervised learning based method including k nearest neighbor or neural network. The method of claim 1, And a preprocessing unit for removing noise with respect to the vibration signal detected by the sensing unit. The method of claim 10, The preprocessing unit is a fall detection device, characterized in that using a high-pass filter (Band-pass filter). The method according to claim 1 or 7, A fall history and a fall are generated by monitoring at least one of an input pattern signaled through the signal processing unit, a fall result of the fall determination unit, a fall result of the post-determination unit, an alarm signal of the warning unit, or a user characteristic stored in advance. Fall detection device further comprises a monitoring unit for notifying the fall state including a point to the outside. Detecting a vibration signal through sensors attached to a floor; Converting the sensed vibration signal from a time domain into a frequency domain and extracting a feature of the vibration signal converted into the frequency domain; Comparing the input pattern corresponding to the extracted feature with a pre-stored standard pattern to determine whether the floor falls; And Falling detection method comprising the step of generating an alarm signal in accordance with the determination result. The method of claim 13, The pre-stored standard pattern is a fall pattern, characterized in that the personalized pattern according to the user-specific body characteristics and the characteristics of the medium of the floor. The method of claim 13, And transmitting the generated alarm signal to a contact point of a pre-stored emergency medical institution or guardian by wired or wireless communication means.

Images