随着社会和技术的发展, 住宅的功能也在变化, 从为人类提供遮蔽的最原始功能, 到如今转变为提供高品质生活的功能。

　　 研究人类的居住行为将有助于高品质的住宅设计, 然而相关研究是现代时期才出现的。从日本家庭的发展变迁与相应生活方式改变的角度, 日本研究者 (Hirai, 1989) 解释了家庭生活、个人行为与居住的关系。陈力等 (1997) 基于行为心理学分析了居民与居住空间的关系^[1], 也有研究者对居住者进行行为监测, 但目的是探讨人类行为对建筑节能性能的影响。在麻省理工学院, 一个多学科的团队 (Intille, 2002) 在为未来的家庭开发技术和设计策略, 他们运用无处不在的传感器, 如基于图像的经验采样和反射, 来定量和定性地研究家庭中人的行为^[2]。

　　 在本研究中, 室内定位系统 (IPS) 被用来连续、实时地定位人们在家中的具体位置^[3]。由于精确的室内定位技术的复杂性, 该领域的研究者已经提出了各种解决方案。其中, 超宽带 (UWB) 是一种新兴技术, 与其他技术相比, 它的性能更好^[4]。UWB是信息传输的一个通道, 它在很大一部分频谱上传播。这允许UWB发射机在消耗少量传输能量的同时能传输大量数据^[5]。Bharadwaj等 (2014) 利用超宽带技术对人体定位进行了研究, 并比较了不同基站数的研究结果, 在使用8个基站时, 研究获得了最小的定位误差。

　　 文章尝试将UWB设备应用于人们居住行为的研究中, 其优势在于它可以连续并且准确地实时追踪人的移动, 希望这种方法可以为未来建筑设计提供一些量化的数据支撑。

　　 UWB是一种信号带宽与中心频率之比大于20%, 或者带宽超过500Mbps的无线射频信号^[6]。UWB发射极短的脉冲, 使用能在很低的功率谱密度下传播无线电能量的技术 (宽频带) 。高带宽为通信提供了高数据吞吐量, UWB脉冲的低频使信号能够有效地穿过障碍物, 如墙壁和目标物。瞬间脉冲有着明显的波峰波谷, 因此信号发射的起止时间更容易测量。这意味着两个UWB设备之间的距离可以通过测量无线电波在它们之间传递所需的时间来被精确计算, 这种技术提供了比用信号强度估计距离更精确的距离测量方法^[7]。

　　 针对UWB开发的最常用算法可分为五大类, TOA (Time of Arrival) 是其中较为常用的方法, 也是本研究所使用的算法^[8]。TOA通过计算多个信号发射方所发射的圆的交集来定位目标。这些圆以信号发射方为中心, 半径是发射方和接收方之间的距离 (图1) ^[9]。由于所有信号发射方的时间是同步的, 接收方可以通过信号的到达时间 (TOA) 来确定它们之间的距离, 即通过信号传播时间乘以光速来计算它们的距离^[10]。

　　 本研究使用的UWB系统, 包括标签、基站、中央处理单元和控制终端以及Nexiot RTLS软件和My SQL数据库。这些标签被做成手环给被测试者佩戴, 基站被安装在每个房间墙面距地面2m高的位置, 它们在一个局域网环境中工作。系统经过多次优化, 数据每2s更新一次, 空间定位精度可达10cm, 可用于各种尺度的分析。

　　 每个标签 (信号发射方) 通过UWB脉冲连续发送信号, UWB脉冲由基站 (参考节点) 接收。然后, 每个基站使用高灵敏度短脉冲检测器来测量信号到达时间。中央处理单元利用接收器的校准数据确定每个信号的到达时间差, 并利用三边算法计算标签的位置。

　　 在我们的长期研究中, 选择了不同家庭结构和住房类型的家庭进行研究, 这里以一对退休夫妇家庭为例。他们居住在一个约77m²的住宅里, 住宅包括两间卧室、卫生间、厨房和客厅, 厨房和客厅由一组立柜隔开。在为期3周的研究中, 我们完成了一系列工作, 包括室内安装调试、连续数据监测、系统优化和问卷调查。设备安装的重点是找到合适的基站安装位置, 以减少信号的衰减、反射及干扰。图2显示了他们的住宅布局和设备安装位置。

　　 经过3周的调查和数据收集, 两位老人的日常活动被记录并存储在MySQL数据库中。我们对数据进行了分类整理, 排除了异常值, 并根据调查问卷等信息对数据进行了校准核对。通过分析数据, 一些常规现象得到了证实, 而有一些现象则为设计提供了新的灵感。这种精确的定位技术可以揭示一些被忽视的居住行为模式和潜在的居住需求。

　　 收集的数据可以从不同的角度来解释。图3展示了这对夫妇日常活动的空间分布, 并描述了他们的空间偏好。可以推断出, 这对夫妇大约早上7点起床, 晚上10点左右睡觉, 每天在13:00~15:00午睡。依据热力图可以看出, 客厅是住宅中利用率最高的空间, 另外值得注意的是, 厨房也是他们生活中一个相对重要的空间。在白天, 数据空白处意味着他们可能在室外, 因为根据UWB定位系统的工作原理, 手环在不移动的时候会进入休眠状态, 长时间的休眠可能是因为他们外出活动了。

　　 当把两位老人的热力图结合在一起分析时, 就会非常清晰地展现出他们的相处方式。例如, 热力图表明他们白天在同一个地方呆的时间比晚上多, 男主人更喜欢呆在厨房, 而女主人喜欢呆在客厅里。这个现象可以通过问卷调查得到解释, 两位老人家有两台电视, 在晚上, 男主人通常喜欢在厨房里看电视, 女主人则喜欢在客厅里看其他电视节目。他们的住宅设计很巧妙, 一方面, 立柜将厨房和客厅分隔, 另一方面, 它又可以让夫妻俩的视线和声音保持联系。

　　 若将一天划分为三个时间段, 上午 (07:00~13:00) 、下午 (13:00~19:00) 、晚上 (19:00~22:00) , 可以看出不同时间段中两位老人在不同房间的停留时长, 以及出入这些房间的频率, 从而可以得到住宅中各个房间的联系强度。如图4所示, 男主人在三个时间段的活动空间以餐厅为主, 中午会在卧室午睡;女主人在三个时间段的活动空间则以客厅为主, 中午会在书房午睡。同时还可以看出, 餐厅和客厅之间的联系强度最强。

　　 以一天为单位的数据则显示了他们在一天内的空间使用偏好 (图5) 。厨房和餐桌在老年夫妇的日常生活中起着重要的作用, 客厅电脑前的空间也是他们的偏好空间。

　　 他们的日常生活非常规律, 这一点从一周内每天的活动频率可以看出 (图6) 。除设备安装时的半天数据空白外, 可以看出他们晚上睡眠时间和中午午睡时间几乎是固定的。而每天他们的白天活动都有一些轻微变化, 男主人活动频率比女主人稍大一些。

　　 通过对上述情况的分析可知, 两位老人喜欢功能复合的空间, 并且希望空间之间能够相互联系、适度连接。因此, 在设计适合诸如退休老人的居住空间时, 应该适当划分空间, 使其更加开放、模糊和多功能, 并注意各个空间之间的联系。

　　 With the development of society and technology, the function of the residential house has changed.Its original function was providing shelter for human.Today, the residence is dedicated to providing people with high-quality life.

　　 Research on the dwelling behaviour is needed for it contributes to good housing design, however, it was only in modern times that we began to study the behaviour of human habitation.From the perspective of the change of Japanese family and the corresponding relation of living form, Japanese researchers (Hirai, 1989) explained the relationship between family life, individual behaviour and dwelling.Chen et al (1997) analysed the relation between dweller and living space based on the behaviour psychology.There are also researches on occupant behaviour monitoring, but the purpose is to explore the impact of human behaviour on building energy saving performance (Yan et al, 2015) .At MIT, a multi-disciplinary team (Intille, 2002) was developing technologies and design strategies for the future home, the ubiquitous sensing such as image-based experience sampling and reflection was used to study human behaviour at home quantitatively and qualitatively.

　　 In this research, indoor positioning systems (IPSs) was used to determine the position of human in their home continuously and in real-time.Due to the complexity of accurate indoor positioning, various solutions have been proposed in this filed.Ultra-wideband (UWB) is an emerging technology that has exhibited better performance compared to others.UWB is a channel of information transmission, which spreads on a large part of the spectrum.This allows the UWB transmitter to transmit large amounts of data while consuming a small amount of transmission energy.Bharadwaj et al (2014) presented a study of human body localization using ultra-wideband technology with various base-stations.The smallest location error has been obtained when using eight base-stations.This paper tries to apply the wearable UWB equipment in the human residential behaviour research.The advantage was it could track the real-time body movement continuously and accurately.Hopefully, it will provide some quantitative basis for future housing design.

　　 UWB is defined as an RF signal occupying a portion of the frequency spectrum that is greater than 20%of the centre carrier frequency, or has a bandwidth greater than 500Mbps.UWB transmits extremely short pulses and uses techniques that cause a spreading of the radio energy (over a wide frequency band) with a very low power spectral density.The high bandwidth offers high data throughput for communication.The low frequency of UWB pulses enables the signal to go through obstructions effectively such as walls and targets.The short signals’bursts, with sharp rises and drops, makes the starts and stops of signals inherently easier to measure.This means that the distance between two UWB devices can be accurately calculated by measuring the time required for a radio wave to pass between them.It delivers much more precise distance measurement than signal-strength estimation.

　　 The most common algorithms which have been developed for UWB can be classified into five main categories.Among them, TOA (time of arrival) and TDOA (time difference of arrival) algorithms are both time-based.TOA locates the target by calculating the intersection of circles for multiple transmitters.These circles are centred on the transmitters, and the radius are the distances between the transmitters and receiver (Figure 1) .As the time of all transmitters is synchronous, the receiver can determine the Time of Arrival (TOA) of the propagating signal from the transmitter and calculate their distance, by multiplying the TOA by the speed of light.

　　 In our research, a set of UWB system was used consisting of tags, anchors, the central processing unit and control terminal including NexiotRTLS software and MySQL database.The tags are worn by people, and the anchors are mounted on the top corners of each room to have line-of-sight with user tag, they work in a LAN environment.The system has been optimized several times, the data is updated every 2 seconds and the spatial positioning accuracy can be up to 10 cm, which can be used for various scales of analysis.

　　 Each tag (transmitter) sends signals continuously with UWB pulse, which will be received by the anchors (reference nodes) .Then each anchor uses a high-sensitivity short-pulse detector to measure the time of arrival of the signal.We have used TDOA for UWB indoor positioning system, the central processing unit will determine the TDOA of each signal with calibration data from receivers and calculate the position of the tag using trilateration algorithm.

　　 In our long-term research, households with different family structures and housing types are selected to study.Here we take a pair of retired couple for an instance.They live alone in a house of about 77m² with two bedrooms, a washing room, a kitchen and a living room, the kitchen and living room are separated by a set of cabinets.During our 3-week research, we completed a series of work, including the indoor installation and debugging, continuous data monitoring, the system optimization, and questionnaire survey.The key point of our work is to find proper locations for the anchors, so as to reduce the attenuation and reflection of signals, as well as the interference.Figure 2 shows the layout of their house and the location of anchors installed.

　　 After 3-week investigation and data collection, the daily activities of the two elderly people are recorded and stored in the MySQL database.We sorted and cleaned the data, excluded the abnormal values, and also checked the data with the water vibration sensors installed.Some phenomena drew from the data are in our expectation, but there are also some beyond our usual thinking.It seems that some ignored residential behaviour mode and potential requirements could be revealed by this precise tracking method.

　　 The data can be interpreted from different perspectives.Figure 3 illustrates the daily activity distribution of the couple and depicts their spatial preferences.It can be inferred that the couple got up at about 7 a.m.and fell asleep around 10 p.m., then took a nap at 13:00-15:00.The spaces with the highest utilization are the kitchen and living room, because they are hottest on the heat map.The kitchen turns out to be a relatively important space in their lives, which attracts our attention.During the daytime, there may be some outdoor activities in the data blanks, because the user tag will go into hibernation when it is not moving, according to how the system works.

　　 When combining the heat map of the host and the hostess together, the pattern how they get along with each other will be clear.For example, it shows that they spend more time in the same place during the day than at night, the host prefers to stay at kitchen while the hostess likes staying in the living room.The explanation was obtained from the questionnaire survey, at night, the host usually watches television in the kitchen and the hostess likes watching other TV programmes in the living room.Their housing design turns to be clever, on one hand, the cabinet separated the kitchen from the living room, and on the other hand, it can keep the couple’s eyesight and sound connected at the same time.

　　 If one day is divided into three periods, morning (07:00-13:00) , afternoon (13:00-19:00) and evening (19:00-22:00) , we can get the stay time of each room in the three periods, and the frequency of access to these rooms, so as to get the connection intensity between each room (Figure 4) .The host mainly stays in the dining room, and he usually takes a nap in the bedroom, however, the hostess stays in the living room at most time, and she usually takes a nap in the study room.As the connection between the rooms, the connection intensity between the dining room and the living room are the strongest.

　　 The whole day’s data shows their spatial preference in one day (Figure 5) .The kitchen and dining table takes up an important part in the old couple’s daily life, and the space in front of computers in the living room is also their preferred space.Their daily routine is very regular, which can be seen from their daily activity frequency within a week (Figure 6) .Except for the data blank of the first half day when the system hasn’t been installed, their sleep time and nap time are almost fixed.However, their daytime activities vary from day to day, and the host seems to be slightly more active than the hostess.

　　 Through the analysis of the above case, people love the space that is functional compound and is moderately connected with other spaces.Therefore, when designing a living space for such families, the space should be appropriately divided to make it more open, fuzzy, and multi-functional, and pay attention to the connection between them.