JP4310409B2 - Life change detection method, apparatus and program - Google Patents

Description

  The present invention relates to a life change detection method, apparatus, and program for detecting a living behavior that is different from a normal resident's behavior.

  In Japan, with the rapid aging of the population, the number of deaths due to accidents in the family has increased rapidly, and in 2001, the number of fatalities in traffic accidents is comparable to that of traffic accidents. In addition, the number of elderly living alone has been on the rise, reaching 3 million in 2000. In such a great change in society, life monitoring technology such as technology for automatically detecting anomalies of consumers and technology for knowing the life of a distant family is required. Therefore, a technology for automatically detecting a state different from usual is being developed based on information obtained from a plurality of sensors arranged in a house.

  For example, Patent Literature 1 discloses a resident's health check method described below. First, current values of a plurality of power supply systems in a house are measured, and feature amounts such as a duration of a constant current value and a current level are set. Then, the data corresponding to each feature amount is input to the neural network, the actual health state is given as a teacher signal, and iteratively calculated until the difference between the output value and the teacher signal becomes sufficiently small while changing each connection weight, Determine each bond load.

    Further, an average value and a standard deviation of the learning feature amount data corresponding to the same time zone for the latest several days are calculated, and virtual data obtained by adding and subtracting a real number multiple of the standard deviation to the average value is created. Then, this virtual data is input as boundary data to the neural network, a teacher signal indicating abnormality is given, and each connection weight is recalculated so that the output signal matches the teacher signal.

After performing the initial setting in this way, feature data is created at several health check timings every day based on the accumulated current value, etc., and when these data are input to the neural network, the health condition is determined. The determination result is output.
JP 2002-109663 A (page 4-8, FIG. 5)

  However, the resident's health check method disclosed in the above-mentioned Patent Document 1 inputs, in an initial setting, feature amount data in a state recognized as a normal health state to a neural network, and a teacher signal representing the health state Therefore, there is a problem that not only is the burden required for the initial setting heavy, but the initial setting must be performed again every time the lifestyle pattern changes, which is troublesome. Furthermore, since it is determined whether the subject is healthy based on the subjectivity of the subject, there is a problem that it is difficult to objectively determine life changes.

  The present invention has been made to solve such a problem, and provides a life change detection method, apparatus, and program capable of objectively detecting a life change without placing a burden on the resident. With the goal.

The object of the present invention is a life anomaly detection method for detecting a living behavior different from normal of a resident, the step of measuring a resident's life state by a plurality of sensors in a time series, and acquiring sensor information; A sensor response pattern at each time by cutting out the sensor information in a predetermined time window for an arbitrary time, and a cluster corresponding to each time is obtained by cluster analysis based on each sensor response pattern. Extracting a plurality of unit time zones having a predetermined time width from the monitoring time zone by the sensor, and based on the transition of the cluster on a plurality of measurement days, the normal time of each unit time zone for each measurement date A step of obtaining a degree, and the obtained ordinary degree is aggregated for a plurality of measurement days, and an average value and a standard deviation of the ordinary degree in each unit time zone are obtained. Determining whether or not there is a lifestyle change by evaluating the normality in the specific unit time zone to be evaluated based on the average value and standard deviation of the normality corresponding to the unit time zone This is achieved by a life abnormality detection method comprising the steps of:

  In this life abnormality detection method, the step of determining the presence or absence of life abnormality normalizes a deviation from an average value of the usual degree in a specific unit time zone to be evaluated by a standard deviation of the usual degree. Accordingly, it is preferable to include a step of calculating a normalized variance score and a step of determining the presence or absence of the life change by comparing the time integral of the normalized variance score with a preset reference value.

  In addition, the object of the present invention is a life anomaly detection device for detecting a living behavior different from a resident's normal, and sensor information for storing sensor information related to the living state of the resident measured by a plurality of sensors Storage means, sensor pattern generation means for acquiring sensor response patterns at each time by cutting out the sensor information in a predetermined time window with respect to an arbitrary time, and cluster analysis based on each sensor response pattern, Cluster analysis means for acquiring a cluster corresponding to the time, and extracting a plurality of unit time zones having a predetermined time width from the monitoring time zone by the sensor, and based on the transition of the clusters on a plurality of measurement dates, A normal degree calculation means for obtaining the normal degree of each unit time zone, and the obtained normal degree is aggregated for a plurality of measurement days, Means for calculating the average value and the standard deviation of the normal degree in the rank time zone, and the normal degree in the specific unit time zone to be evaluated, the average value of the normal degree corresponding to the unit time zone And it is achieved by a life change detection device comprising a life change determination means for determining the presence or absence of a life change by evaluating based on the standard deviation.

  The object of the present invention is achieved by a life change detection program for causing a computer to function as the life change detection device.

  According to the present invention, it is possible to detect life changes objectively without imposing a burden on the resident.

    Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings. FIG. 1 is a block diagram showing a life change detection device according to an embodiment of the present invention. As shown in the figure, this life change detection device 1 is connected to a plurality of sensors S1, S2,..., Sn, and includes sensor information storage means 2, sensor pattern generation means 4, cluster analysis means 6, An ordinary degree calculating means 8, an ordinary degree analyzing means 9, and a life change judging means 10 are provided. The plurality of sensors S1, S2,..., Sn are preferably those that can be constantly monitored without disturbing the living state of the resident, and examples thereof include an infrared sensor and an electric energy sensor. Further, the number of sensors is not particularly limited as long as it is plural, but it is preferable to arrange the sensors so that the movements of the residents in the entire house can be grasped. Input of sensor information to the life change detection device 1 can be performed by wireless communication, for example.

The sensor information storage means 2 stores measurement data that is constantly input from a plurality of connected sensors S1, S2,..., Sn as sensor information. The sensor pattern generation means 4 acquires a sensor reaction pattern at each time based on the sensor information. The cluster analysis means 6 acquires a cluster corresponding to each time by cluster analysis based on the sensor reaction pattern.

    The usual degree calculation means 8 extracts a plurality of unit time zones having a certain time width from the monitoring time zones by the sensors S1, S2,..., Sn, and measures based on the transition of clusters on a plurality of measurement days. The usual degree of each unit time zone is calculated every day. The ordinary degree analyzing means 9 aggregates the ordinary degree obtained by the ordinary degree calculating means 8 for the same unit time zone on a plurality of measurement days, and calculates the average value and standard deviation of the ordinary degree in each unit time zone. The life change determination means 10 determines the presence or absence of a life change by evaluating the normality in a specific unit time zone to be evaluated based on the average value and standard deviation of the normality in the same unit time zone.

In the present embodiment, the cluster analysis unit 6 performs component analysis such as principal component analysis on each sensor reaction pattern, thereby extracting feature information consisting of a plurality of main components, and each sensor And feature amount generation means 62 for obtaining feature amounts such as principal component load amounts corresponding to each time based on the feature information from the reaction pattern, and by performing cluster analysis on this feature amount, the cluster at each time is obtained. Configured to get.
In the present embodiment, the normality calculation means 8 includes standard cluster information generation means 81 for generating standard cluster information based on cluster transitions on a plurality of measurement dates, and the cluster on a specific measurement date to be evaluated. Specific cluster information generation means 82 for generating specific cluster information based on the transition of the standard cluster information, and the usual degree is calculated based on the standard cluster information and the specific cluster information.

Moreover, in this embodiment, the life abnormality determination means 10 is a normalized variance score obtained by normalizing the deviation from the average value of the normal degree in the specific unit time zone to be evaluated by the standard deviation of the normal degree. The normalization variance score calculation means 101 is calculated, and the presence / absence of life change is determined by comparing the time integral of this normalized variance score with a preset reference value.
Next, an embodiment of a method for detecting a life change using the life change detection apparatus 1 described above will be described. In this embodiment, a case where an in-room sensor and a power sensor are used as the sensors S1, S2,..., Sn and installed in a house of a family of four (a couple and two children) will be described. As each installation place and measurement object, what is shown in FIG. 2 can be mentioned, for example.
As the occupancy sensor, one having a configuration in which four infrared sensors having detection areas in four different directions are attached to the lower part of the lighting fixture and one having a configuration having one infrared sensor are emitted from the human body. It can be configured to detect an infrared intensity change and transmit an ON / OFF signal every second. About what is mounted | worn with a lighting fixture, it can comprise so that the ON / OFF signal of a lighting fixture may also be transmitted. Moreover, the electric energy sensor can be configured to output an integrated value of electric power used by various connected home appliances at intervals of 1 minute.
After such sensors S1, S2,..., Sn are connected to the life abnormality detection device 1 so as to be capable of wireless communication, time series measurement is started. The measured data is stored in the sensor information storage unit 2 as time-series information of sensor information. An example of sensor information is shown in FIG. The sensor information is stored for a predetermined period (for example, about two weeks).
The time-series information shown in FIG. 3 is information that changes at intervals of 1 second at the shortest, and the sensor response is different even in the same living state (a state such as a meal or relaxation). Is difficult to detect. Therefore, in the present embodiment, the cluster analysis unit 6 performs cluster analysis based on the sensor reaction pattern generated by the sensor pattern generation unit 4 to convert the sensor information into information representing a living state.

More specifically, the sensor pattern generation unit 4 generates sensor response patterns by cutting out sensor information through a time window w corresponding to the measurement start time. The time window w is set to 10 seconds, and sensor response patterns are sequentially acquired while moving the time window every second. Feature information extraction means 61
Performs principal component analysis of each sensor response pattern obtained (see FIG. 3).

The principal component analysis method can be performed using various known algorithms. In this embodiment, the sensor response pattern per day obtained every second in a time window of 10 seconds is about 86400 patterns. Since the number is enormous, it is difficult to obtain the principal component of sensor information over several weeks by a direct solution method that calculates eigenvalues. Therefore, a generalized Hebbian algorithm for obtaining principal components using a learning function of a neural network can be preferably exemplified. According to this algorithm, when obtaining n principal components, the i-th principal component vector c _i (i = 0,..., N−1) is the sensor response pattern x after k seconds cut out by the time window w. By sequentially inputting ^(k) , the following equation 1 is obtained recursively.

ただし、γは定数で、左肩添え字ｔはベクトルの転置を表す。
主成分の数ｎは、寄与率の総和が所定値（例えば０．８程度）を超える最小の値とすることにより自動的に設定可能であり、これによって、センサ反応の中の主要な成分のみを抽出することができ、複数の主成分から構成される特徴情報を得ることができる。図４は、１４日間のセンサ情報から求めた特徴情報の一例を示しており、主成分１〜主成分１０までの１０個の主成分により構成されている。

However, γ is a constant, and the left superscript t represents transposition of the vector.
The number n of the main components can be automatically set by setting the sum of the contribution ratios to a minimum value exceeding a predetermined value (for example, about 0.8), whereby only the main components in the sensor reaction can be set. And feature information composed of a plurality of principal components can be obtained. FIG. 4 shows an example of the characteristic information obtained from the sensor information for 14 days, and it is composed of 10 main components from the main components 1 to 10.

Next, the feature quantity generation means 62 calculates principal component load quantities (p _i = ^t c _i x / | c _i |) obtained by projecting the sensor reaction patterns cut out in the time window w onto the respective principal component vectors at each time. Ask for it. Cluster analysis is performed on this principal component loading.

The cluster analysis method is also not particularly limited, and examples thereof include a method by ART (Adaptive Resonance Theory) that can generate a non-hierarchical cluster in a self-organizing manner. That is, if the vector p having the principal component load amount as an element is different from the direction of the already generated m clusters q _i (i = 0,..., M−1), that is, all q _A new cluster p is generated when the following equation 2 holds for _i .

図５は、図４に示す１０個の主成分に基づく主成分負荷量のクラスタ分類（但しα＝０．６）により生成された生活情報クラスタの一例であり、クラスタ１〜クラスタ４４までの４４種類のクラスタが示されている。尚、クラスタ分類された際の各主成分負荷量に対する重みづけが横軸に並んでおり、図中の横軸の数字は各主成分の番号に対応している。これらのクラスタは、「食事」や「くつろぎ」などの生活の意味との関連性が必ずしも強い訳ではないが、その家族における主要な生活状態を表現していると考えられる。参考として、センサ情報を主成分分析することなく直接クラスタ分析したところ、図６に示すように、細かなセンサ反応に対応したクラスタが多数生成され、生活状態を記述するには不都合な結果となった。この結果から、主成分負荷量をクラスタ分析する有効性が確認できる。
標準クラスタ情報生成手段８１は、各時刻に対応するクラスタにより、日々の生活をクラスタ間の遷移パターンとして表す。図７は、遷移パターンの一例を示しており、各日付におけるクラスタ間の遷移状態を表している。この結果から、複数日の同時刻におけるクラスタに基づいて、当該時刻のクラスタ発生頻度分布を作成する。ある時刻におけるクラスタの発生頻度の最大値が１になるように正規化すると、図８に示すように、１日の各時刻におけるクラスタの発生頻度分布に基づく、その家族の標準生活テンプレート（標準クラスタ情報）が得られる。尚、図８において、発生頻度が高いところほど明るく表示されている。特定クラスタ情報生成手段８２は、評価対象となる特定の日のクラスタ遷移パターン（特定クラスタ情報）を生成する。

FIG. 5 is an example of the life information cluster generated by the cluster classification (α = 0.6) of the principal component load amounts based on the ten principal components shown in FIG. Types of clusters are shown. In addition, the weighting with respect to each principal component load amount at the time of cluster classification is arranged on the horizontal axis, and the numbers on the horizontal axis in the figure correspond to the numbers of the respective principal components. These clusters are not necessarily strongly related to the meaning of life such as “meal” and “relaxation”, but are considered to represent the main living conditions in the family. As a reference, when cluster analysis is directly performed on the sensor information without performing principal component analysis, as shown in FIG. 6, many clusters corresponding to fine sensor reactions are generated, which is inconvenient for describing the living state. It was. From this result, the effectiveness of cluster analysis of the principal component load amount can be confirmed.
The standard cluster information generating unit 81 represents daily life as a transition pattern between clusters by a cluster corresponding to each time. FIG. 7 shows an example of the transition pattern, and represents the transition state between clusters on each date. From this result, based on clusters at the same time on a plurality of days, a cluster occurrence frequency distribution at that time is created. When normalization is performed so that the maximum value of the occurrence frequency of a cluster at a certain time becomes 1, as shown in FIG. 8, the standard life template (standard cluster) of the family based on the occurrence frequency distribution of the cluster at each time of the day as shown in FIG. Information). In FIG. 8, the higher the occurrence frequency, the brighter the display. The specific cluster information generation unit 82 generates a cluster transition pattern (specific cluster information) on a specific day to be evaluated.

  The ordinary degree calculation means 8 evaluates the ordinary degree in each unit time zone of each measurement day by comparing the standard life template composed of the above-described cluster occurrence frequency distribution with the cluster transition pattern of the day to be evaluated. .

The probability of occurrence of cluster c at time t represented by the standard life template is P (t; c)
When the cluster c appears at time t, X (t; c) = 1, and when the cluster c does not appear at time t, X (t; c) = 0, the time width T at time t is The evaluated ordinary degree O _T (t) is given by the following equation 3.

普段度Ｏ_Ｔ(t)は、標準生活テンプレートの同じ単位時間帯で示される最も発生しやす
いクラスタだけが起きた場合に最大値１を示し、値が小さくなる程、普段の生活状態から逸脱が大きいことを示している。時間幅Ｔを１日とし、開始時刻τが１秒ずつずれるように単位時間帯を設定した場合における普段度の変化の一例を図９に示す。昼間は様々な生活状態が発生するため普段度が低くなる一方、夜は通常睡眠の状態であるため普段度が高くなっている。こうして得られた各測定日の各単位時間帯における普段度は、メモリに格納される。

The usual degree O _T (t) shows a maximum value of 1 when only the most likely cluster shown in the same unit time zone of the standard life template occurs, and the smaller the value, the more the deviation from the normal life state. It is big. FIG. 9 shows an example of a change in the normality when the unit time zone is set such that the time width T is 1 day and the start time τ is shifted by 1 second. While various living conditions occur during the daytime, it is usually low, while at night it is normal because it is a normal sleep state. The usual degree in each unit time zone of each measurement date thus obtained is stored in the memory.

The ordinary degree analysis means 9 totals the obtained ordinary degrees and calculates the average value and standard deviation of the ordinary degrees in each unit time zone. FIG. 10 shows an example of the average value and the standard deviation of the usual degree when the time width T of the unit time zone is 1 hour. Since the state is stable due to sleep at night, the average value is usually high and the standard deviation is small. On the other hand, since the living state is not constant during the daytime, the average value of the usual degree becomes low and causes variation. However, there is a time zone in which the same living state is likely to occur even during the daytime depending on the time zone, and the standard deviation is small in this time zone.
The normalized variance score calculation means 101 calculates a normalized variance score according to the following equation 4 from the usual degree in a specific unit time zone on a specific day to be evaluated.

こうして得られた正規化分散得点は、平均値が普段度の平均値と同じであり、標準偏差がどの単位時間帯でも１となるように正規化されている。すなわち、図１１に示すように
、時刻ｔ１及びｔ２における普段度が同じであっても、時刻ｔ１における標準偏差は時刻ｔ２における標準偏差よりも小さいために、時刻ｔ１の正規化分散得点は時刻ｔ２の正規化分散得点よりも小さくなる。このような指標を用いることにより、単に普段度のみで評価する場合に比べて、生活異変の状態をより正確に表すことができる。

The normalized variance score obtained in this way is normalized so that the average value is the same as the average value of the usual degree and the standard deviation is 1 in any unit time zone. That is, as shown in FIG. 11, even if the usual degrees at times t1 and t2 are the same, the standard deviation at time t1 is smaller than the standard deviation at time t2, so the normalized variance score at time t1 is the time t2. Smaller than the normalized variance score. By using such an index, it is possible to more accurately represent the state of life change than in the case of simply evaluating based on the usual degree.

  The life anomaly determination means 10 calculates the evaluation score p (t) by time-integrating the obtained normalized variance score based on the following equation 5.

評価得点は、正規化分散得点の負の値のみを積分することにより求められ、正規化分散得点が正の値をとる時は、積分値が０にリセットされる。こうして得られた評価得点は、生活異変の程度が大きく、また異変の時間が長いほど、負の値が大きくなるため、この評価得点を予め設定された基準値（例えば、正規化分散の標準偏差値のｎ倍を負にした値）と比較し、評価得点が基準値よりも低い場合を生活異変として検知することができる。

The evaluation score is obtained by integrating only the negative value of the normalized variance score. When the normalized variance score takes a positive value, the integral value is reset to zero. The evaluation score thus obtained has a greater negative value as the degree of life change is longer and the time of the change is longer. Therefore, this evaluation score is set to a preset reference value (for example, standard deviation of normalized variance). When the evaluation score is lower than the reference value, it can be detected as a life anomaly.

  FIG. 12 shows the normalized dispersion score obtained from monitoring the actual living state and obtained from the measurement data for one month from April 1 to April 30, and the evaluation score based thereon. FIG. 13 shows an unusual life state reported by a consumer.

  As shown in FIG. 12, the unit time zone determined to be a life anomaly based on the evaluation score calculated from the normalized variance score is generally consistent with the report contents of the consumer, and the actual life anomaly is well Reflects. As described above, according to the life change detection device of this embodiment, it is possible to accurately detect a life change without placing a burden on the resident.

  In addition, even if the floor plan of the house, sensor arrangement, family members, etc. change, it can be automatically reflected in the standard life template (that is, standard cluster information such as cluster occurrence frequency distribution) by accumulating sensor information. Because it can, it can easily cope with environmental changes.

  In addition, by storing the average value and standard deviation of usual times in advance, it is possible to quickly detect changes in life by setting the time width of the unit time zone to an appropriate value (for example, about 1 hour). And can effectively report to external organizations.

  In the present embodiment, as an example of the standard life template (standard cluster information), a case where a standard cluster pattern including a cluster occurrence frequency distribution using an occurrence frequency (occurrence probability) of a cluster as an evaluation index has been described. It is also possible to use other evaluation indices such as transition probability, duration probability, period probability, and inter-cluster transition frequency, and by combining each evaluation result using these multiple evaluation indices, It can also be determined.

    In the above embodiment, the cluster analysis means 6 performs principal component analysis on the sensor response pattern to extract feature information, and the principal component load amount is obtained from each sensor response pattern based on the feature information. Although it is configured, the component analysis method is not necessarily limited to principal component analysis, and factor analysis, independent component analysis, and the like can also be applied.

    It is also possible to create a program for causing a general-purpose computer to function as the above-described life change detection device and distribute the program in the market.

It is a block diagram which shows the life abnormality detection apparatus which concerns on one Embodiment of this invention. It is a figure which shows one Example of the installation place of the sensor connected to the said life abnormality detection apparatus, and a measuring object. It is a figure which shows one Example of sensor information. It is a figure which shows one Example of characteristic information. It is a figure which shows one Example of a cluster. It is a figure which shows the comparative example of a cluster. It is a figure which shows one Example of the transition pattern between clusters. It is a figure which shows one Example of a standard life template. It is a figure which shows one Example of usual times. It is a figure which shows one Example of the average value and standard deviation of usual degree. It is a figure for demonstrating the relationship between usual degree and a normalization dispersion | distribution score. It is a figure which shows one Example of a normalization dispersion | distribution score and an evaluation score. It is a figure which shows the report content of a consumer's life state.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Life change detection apparatus 2 Sensor information memory | storage means 4 Sensor pattern production | generation means 6 Cluster analysis means 61 Feature information extraction means 62 Feature quantity generation means 8 Normal degree calculation means 81 Standard cluster information generation means 82 Specific cluster information generation means 9 Normal degree analysis Means 10 Life change determination means 101 Normalized dispersion score calculation means S1, S2,..., Sn sensor

Claims (8)

A method for detecting abnormal life in order to detect a resident's daily living behavior,
Measuring a resident's life state with a plurality of sensors in time series, and obtaining sensor information;
Obtaining a sensor response pattern at each time by cutting out the sensor information in a predetermined time window with respect to an arbitrary time; and
Obtaining a cluster corresponding to each time by cluster analysis based on each sensor response pattern;
Extracting a plurality of unit time zones having a predetermined time width from the monitoring time zone by the sensor, and determining the normality of each unit time zone for each measurement day based on the transition of the cluster on a plurality of measurement days; and ,
Aggregating the obtained ordinary degree for a plurality of measurement days, calculating an average value and a standard deviation of the ordinary degree in each unit time zone,
Determining whether or not there is a life change by evaluating the normality in the specific unit time zone to be evaluated based on an average value and a standard deviation of the normality corresponding to the unit time zone. Life change detection method. The step of determining the presence or absence of the life change includes
Calculating a normalized variance score by normalizing a deviation from an average value of the usual degree in the specific unit time zone to be evaluated by a standard deviation of the usual degree;
The life change detection method according to claim 1, further comprising: determining whether or not there is a change in life by comparing the time integral of the normalized variance score with a preset reference value. Obtaining the cluster comprises:
Extracting component information consisting of a plurality of components by performing component analysis on each sensor response pattern; and
Obtaining each feature quantity corresponding to each time based on the feature information from each sensor response pattern;
The lifestyle abnormality detection method according to claim 1, further comprising: obtaining a cluster at each time by performing cluster analysis on the feature amount. The life abnormality detection method according to claim 3, wherein the component analysis is a principal component analysis, and the feature amount is a principal component load amount. The step of obtaining the usual degree is:
Generating standard cluster information based on transitions of the clusters on multiple measurement dates;
Generating specific cluster information based on transition of the cluster on a specific measurement date to be evaluated;
The lifestyle change detection method according to claim 1, further comprising a step of calculating the normality of each unit time zone based on the correlation between the standard cluster information and the specific cluster information. The life abnormality detection method according to claim 1, wherein the plurality of sensors include a plurality of infrared sensors and a plurality of electric energy sensors. It is a life anomaly detection device for detecting a living behavior different from a normal resident,
Sensor information storage means for storing sensor information relating to the living state of the resident measured by a plurality of sensors;
Sensor pattern generation means for acquiring a sensor response pattern at each time by cutting out the sensor information in a predetermined time window with respect to an arbitrary time;
Cluster analysis means for obtaining a cluster corresponding to each time by cluster analysis based on each sensor response pattern,
A normal degree of extracting a plurality of unit time periods having a predetermined time width from a monitoring time period by the sensor, and obtaining a normal degree of each unit time period for each measurement day based on a transition of the cluster on a plurality of measurement days A calculation means;
The obtained ordinary degree is aggregated for a plurality of measurement days, and an ordinary degree analyzing means for calculating an average value and a standard deviation of the ordinary degree in each unit time zone,
Life abnormality determination means for determining the presence or absence of a life abnormality by evaluating the normality in the specific unit time zone to be evaluated based on the average value and standard deviation of the normality corresponding to the unit time zone A life change detection device comprising: A life change detection program for causing a computer to function as the life change detection device according to claim 7.

Images