• Definition

Thermal comfort as we refer to in the field of building physics is traditionally defined through people's psychophysical responses, rather than purely biophysical processes. Thermal comfort is a subjective state of mind where each individual, influenced by physical, physiological and psychological factors, expresses a judgement of satisfaction with the thermal environment [14,47,49]. While thermal comfort is a highly subjective condition, thermal sensation is more objective and is therefore used to describe the human response to thermal comfort [146]. Other indicators used include thermal acceptability (the degree of an occupant's approval of the environment, which is subjective and directly related to the individual's expectation) and thermal preference (the expressed ideal thermal state of the environment) [30]. Variables that influence the comfort sensation are multiple. On one hand are the physical factors defining the thermal state of the environment: mean radiant temperature, relative humidity, air velocity and air temperature [14]. On the other are the factors influencing the human perception and preference towards the thermal conditions: individual factors such as age, gender sex, metabolism rate, mood, etc. [34,147]; dynamic factors such as clothing, activity patterns, posture (sedentary or steady conditions) [58]; and contextual climatic conditions as geographical factors, weather, and time of the year [14].

• Thermal Comfort Models

Achieving optimal indoor thermal comfort conditions is a complex task that has been at the center of scientific debate for about fifty years [50,62]. The outcome has been two approaches for defining thermal comfort.

The "single temperature optimum" model is based on climate chamber data, on the heat balance theory and on thermoregulation physiology [59]. This analytical model defines the predicted mean vote (PMV) index and the predicted percentage of dissatisfied occupants (PPD) through physical parameters (air temperature, mean radiant temperature, air velocity and relative humidity) and human variables (clothing insulation and activity level) [62]. Thermal comfort is therefore expressed around a single optimum temperature for any given combination of comfort parameters [148]. In this approach, any deviation from the thermal optimum anticipates a loss in comfort, performance, and productivity [51].

The "adaptive comfort" model on the other hand is based on field studies and considers that occupants' thermal acceptability, which depends on behavioral, physiological and psychological factors, influences the thermal comfort perception [14,49]. While subjects in lab conditions expect a finely controlled thermal environment, calculating the comfort predictions to a pessimistic value and are therefore less tolerant [50], in environments dependent on outdoor temperature fluctuations and passive design solutions, users accept a broader thermal comfort range [50,51]. Higher comfort and cognitive demands are also expected to be absorbed to a certain degree by the subjects with little or no deleterious effect appearing until those adaptive resources are depleted [51].

While both the "single temperature optimum" and the "adaptive comfort" models rely on thermal behavior to a certain extent, the thermoneutral zone (TNZ) describes a range of ambient temperatures where the body's core temperature can be maintained as relatively constant solely through control of dry heat loss (without regulatory changes in metabolic heat production or evaporative heat loss) [46,149]. The concept distinguishes itself from the thermal comfort models [149] as being within the thermal comfort zone does not guarantee that the body is in thermal balance at basal metabolic rate, or that it does not require heat production or heat loss to maintain the core temperature [46].

In all three concepts, the physical parameters can be measured using sensors. Subjective parameters, however, are more diverse and different methods are used to collect human responses, from complaint analyses, online surveys [14] and/or self-report. While the subjective measurement has improved greatly across several disciplines over the years, the soft data still pose problems in regard to being statistically significant, reproducible [56], or scalable.

#### 3.1.3. Visual IEQ

Visual IEQ is affected by parameters such as daylight (in both amount and quality), the quality of views (indoors, towards the outside, introspection) and the user's opportunities to adjust these [133].

• Light

Visual comfort is generally the primary aspect considered by designers when planning the indoor light conditions and is in many cases the only aspect measured to determine visual IEQ. Visual comfort is determined by daylighting and artificial lighting levels in both amount and quality: overall luminance levels, daylight to artificial light ratio, direct sunlight, glare index, etc. It additionally depends on the type of activity performed, as different activities require different lighting levels [150].

Exposure to natural daylight is however important for more than its optical properties [44]. The intensity of the light affects human health as well as exposure to specific wavelengths, the timing, and the duration of exposure [151]. Humans have evolved by being exposed to specific amounts of natural sunlight. Spending more time indoors can therefore result in too low exposure, especially as window glasses tend to block part of the natural light spectrum [152]. Indoor sunlight can therefore not entirely substitute outdoor natural light. In terms of intensity, natural light in a minimum of 1000 lux is required for the biological rhythms of the human body to work efficiently [14], while most indoor office lighting levels for instance are around 300–500 lux [44]. In terms of wavelength, spending too much time indoors can lead to low exposure to ultraviolet (UV) radiation, which involves vitamin D3 deficiency [10]. In terms of timing, longer exposures to artificial light enriched in blue, typically irradiated from electronic screens, have been shown to affect sleep structure, particularly when exposed before bedtime [153]. Ill-timed light exposure due to the use of electrical light during periods of the day with local environmental darkness (i.e., late evening, night, early morning) has been shown to result in melatonin suppression even in settings with low light levels such as in households [10]. Melatonin is a hormone responsible for regulating the body's circadian rhythm [154,155] on which human physiological and performance aspects depend [154,156]. A multitude of disorders can emerge from the disruption of the circadian rhythm, such as sleep [153] and mood dis-

orders [157], seasonal affective disorder (SAD) [10] and shift work disorder (SWD), which create issues such as insomnia, difficulty in falling asleep, and experiencing sleepiness when it is important to be alert and productive [157].

Indoor lighting levels are mostly measured using sensors, while the glare index is mostly assessed through calculations [14]. Preferred indoor lighting settings can be assessed by comparing the lighting level and glare with occupant survey data [152].

In summary, relevant light-connected parameters to ensure good IEQ include daylight and artificial parameters in terms of quantity and quality: overall luminance levels, daylight/artificial light ratio, direct sunlight, glare index, wavelength, timing and the duration of exposure.

• Views

The quality of views impacts occupants' health and productivity in a number of ways [52]. Within the indoor environment, aesthetics and color schemes have been shown to affect human performance and productivity [14,158]. Greenery, whether in direct contact with the users [159], or indirectly spotted through windows [39,160] has therapeutic psychological and physiological restorative effects such as stress reduction, development cognitive and social skills [79]. Viewing the outdoors does not only provide contextual information such as weather, nature and surrounding activities [14].

#### 3.1.4. Acoustic IEQ

Acoustic IEQ has high relevance in building design, and especially in offices as many tasks require noise to be kept within certain limits to enable occupants to work efficiently. Bad acoustic conditions produce psychological annoyance [161,162], fatigue and negative impact on motivation [31], anxiety with increased stress levels [163], ultimately affecting users' performance [32] and creating long term health issues [14].

Any unwanted sound is referred to as "noise" [164]. Sources of noise include external sources (traffic, the public, air traffic, machinery), internal sources such as machinery (fax machines, telephones, air conditioning systems) or from human origin (co-worker conversations, one-minute requests, etc.) [133,161].

Sound (or noise) is measured on a logarithmic arithmetic scale of decibels (dB), sound power (SWL) and sound pressure levels (SPL) [14]. The effect of variations in acoustic sensations in users' productivity has been directly compared to those of changes in thermal sensation, whereas a temperature change of 1 ◦C supposedly has the same effect as a change in noise of 2.6 dB [61]. The typical sound range in an air-conditioned office is between 45 dB and 70 dB [165,166].

Reducing acoustic discomfort can be achieved in a number of ways depending on the source of noise and the context: adding external (building envelope) [14] or internal building elements (internal partitions, sound absorbing materials, modifying sound reverberation time) [167], modifying the internal layout (from open office plans to cubicle/cellular office arrangements) [14], using a white noise generator [168] and introducing vegetation to promote reflection, dispersal, absorption or interference with the sound waves [78,169].

#### *3.2. Biosignals*

3.2.1. Definitions and Classifications

Living organisms, depending on their complexities, are made of several dynamic biologic systems. In human beings, some of these systems can be listed as the nervous system, cardiovascular system, musculoskeletal system, or the immune system [170]. These systems are responsible for dedicated physiological processes, such as blood circulation, breathing, digesting, etc. These processes result in changes within themselves, their immediate environment and/or their input/outputs in forms of voltage, pressure, chemical concentration, temperature, etc. These physical attributes can be measured by several means and are collectively called biosignals.

Biosignals (short for biologic signals or biomedical signals) are signals that are used to extract information on a biologic system to be examined [171]. In biomedical applications, they are a critical part of diagnosis [172]. The signal source can be at the molecular, cell, systemic or organ level. Extraction of a biosignal can go from something as simple as a physician feeling the pulse of a patient to understand the heart rate, to the use of an electrocardiogram or to measure the electrical activity of the heart more precisely and continuously, with the help of electrodes placed on the chest, arms, and legs. Overall, biosignals can be:


Biosignals can be categorized in several ways: by their existence (permanent, i.e., EEG, ECG vs. induced, i.e., plethysmography, where an artificial current is induced in the tissue), by the observed time-frame changes (dynamic, i.e., heart rate or static, i.e., core temperature) or by their physical nature [174]. The latest is the most frequent classification method, categorized and elaborated as follows:


7. Bio-optical signals: Bio-optical signals are naturally occurring or induced optical functions of the examined biologic system. Examples of use include estimating blood oxygenation by measuring transmitted vs. backscattered light from a tissue, using dye dilution and monitoring the bloodstream to observe cardiac output, or controlling fluorescence characteristics of the amniotic fluid to acquire information about the health of the fetus [171].

All of these types of biosignals have different levels of sensitivity, accuracy, timescale, and types of dedicated sensors [183]. The use of these sensors suggests another important classification relevant to the focus of this research: the level of invasiveness. This factor is mainly critical in order to eliminate any psychological bias on one hand, by ensuring that the test subjects are not aware of the experiment, but also relevant due to the applicability of the biosensing experiments outside laboratory environments on the other.

The level of invasiveness is a direct correlation between the biosignal sensor and its relation to its biologic host; naturally, the least invasive biomedical data collection methods would be through the use of no contact sensors at all (i.e., thermal infrared imaging). Minimally invasive methods can be directly in contact with the skin (i.e., measurements obtained through biopotential electrodes such as electrocardiogram, electroencephalography, electrodermal activity), whereas the most invasive methods would implement sensors within the body (i.e., rectal, oral thermometers), puncturing the skin (i.e., use of needle electrodes for electromyographic signal acquisition) or even surgically placing the sensor (i.e., a blood pressure sensor placed in an artery, vein, or in the heart) [29].

#### 3.2.2. Use of Biosignals in the Field of Building Engineering

The human body constantly tries to self-regulate against environmental changes, trying to keep the state of homeostasis and, as defined by Selye [184], stress is "a state of biological activation triggered by the person interacting with external agents that force her or his capacity to adapt". According to Schneiderman et al. [130], following a stressful event, a cascade of changes in the nervous, cardiovascular, endocrine, and immune systems take place. As previously discussed, these changes are traceable and quantifiable through sensing of biosignals.

Stress literature is well researched and has a long history. From a medical point of view, it goes back to the beginning of the 20th century [185] and gained even more traction in fundamental and clinical neuroscience research in the 50s, after Hans Selye's stress theory [186]. Since then, stress has been a research topic in many specific fields such as psychology, economics, ergonomics, sociology, endocrinology, complementary medicine, animal breeding, etc. It has been a research topic in the domain of building physics as well, regarding occupant stress (mental or physical, in terms of *thermal stress, heat stress* or *cold stress*), as reviewed in the first chapter.

However, the use of biosignals as part of stress-research is relatively new, even more so in the field of building engineering. As indicated in Section 2, Materials and Methods, the list of publications researched for this review mostly tried to correlate IEQ parameters to comfort sensation and mental workload, while none, to our knowledge, investigated a long-term effect.

#### 3.2.3. Limitations

Albeit promising, the use of biosignals in the field of building engineering has still several limitations, most of which pose setbacks for robust real-world applications so far. The majority of the studies using biosignals have been conducted in test chambers, laboratories or well-controlled environments, and the gap in research to transfer the knowhow from the lab setting to real-world conditions still exists [120]. While this approach is beneficial in establishing working protocols with minimal noise and more comparable data, it is also well-known that people's tolerance limits may drastically vary from test environments to real-world settings in which there will be many unforeseen stressors, influencing the overall stress [122,187]. However, both the sensing and data analysis

technologies are rapidly developing, and the possibility of the use of biosignals to further develop our know-how in realistic settings does not seem unrealistic.

The following are the most critical obstacles, according to our literature review.

	- a. Noise:

The physiological processes that produce biosignals, most of the time, cannot be reduced to an unequivocal course of events. Motion of the entire body, or parts of it, creates noise in the data. While there are several methods to improve signal-to-noise ratio, and higher performing sensors already eliminate noise to a certain degree, this still poses a problem. While de Santos et al. [188] remarks that "the subjects were strongly indicated not to move during the experiment procedure, in order to avoid noise in physiological signal acquisition", it significantly deviates from a realistic setting, ergo realistic conclusions, since it not only is irrational to expect test subjects to be that restricted while doing their mundane tasks, it is also effective on the strain this command causes on the subjects, meaning the results will not be representative for real-world settings.

Similarly, Schmidt et al. [189] states that certain biosignals, such as electrooculography or electromyogram, are prone to producing a lot of noise in real world settings, and therefore are not very suitable for any use outside the laboratory conditions.

b. Level of invasiveness, mobility, or wearability of sensors:

The last decade has seen a great development in wearable technologies. The wearables used in the field can be watch-like, chest-belt, stationary devices, or recently flexible sensor patches and sensors that can be integrated into fabric. These technologies offer an increased wearing comfort, potential new measurement positions, and are less intrusive [189]. Even biochemical signals such as saliva and sweat, which are sought after as direct stress measurements, can be analyzed non-invasively through recently developed wearable electrochemical sensors [190].

However, there are still problems associated with mobile technologies: once stationary devices, EEGs now have mobile alternatives with wearable headsets, with wireless data transfer technology. Nevertheless, they are still not fully unobtrusive; in some experiments, it was reported that continuous wearing of said headset caused headaches [2], making it less suitable for prospective long-term experiments in real settings.

c. Accuracy of sensors:

Another topic with mobile and wearable sensors as an emerging concern has been the level of accuracy that can be obtained during the testing period [191]. Continuing with the example of mobile EEGs, mechanical setbacks still exist; several studies [192,193] have explored the drop in signal quality and accuracy after a certain time of usage, due to sweat on the skin changing the bioimpedance of electrodes.

2. Problems with data acquisition:

Depending on country-specific regulations, data privacy is strictly controlled and regulated. This means that data collection methods, particularly when video and/or speech recording is involved, might have to be modified or restricted accordingly [29].

3. Need for self-reporting:

Sharma and Gedeon [194] and Hernandez et al. [195] state that most methods that use biosensing techniques (alone or in combination) still use questionnaires to validate induced stressors. This eliminates the potential avoidance of psychological bias in the resulting data. On the other hand, Hernandez et al. [196] suggests that self-reports, when used as ground-truths, result in more accurate stress detection in person-specific models.

4. Need for multi-modal biosignals for better insight:

Generally speaking, multimodal systems perform better when compared to the accuracy reached by unimodal systems [197]. However, Kyriakou et al. [120] make a critical

point by stressing the importance of combining the appropriate signals and signal processing algorithms. A multimodal data collection is also practical in order to study the cross-modal effects that one can expect to observe in a real-world setting [198].

Thereby also increasing the applicability problems parabolically, making it even more complicated under the real-world conditions.

#### 3.2.4. State of the Art

In this chapter, relevant biosignals used to assess environmental stressors in the built environment are reviewed and classified. While there is a plethora of biosignals, the ones presented in this chapter are selected on the basis of their occurrence in stress studies directly related to IEQ parameters.

The presented biosignals are simply classified regarding their position in the human body.

• Brain: Electroencephalogram (EEG)

Since the early 20th century [199,200], brain activities can be recorded by electrodes. The system is called electroencephalogram, or EEG for short; by "electro = electrical", "encephalo = brain" and "gram = record" and used for brain function study [201]. EEG is the electrical recording of brain activity, represented as voltage fluctuations resulting from ionic current flows within the neurons of the brain [202]. EEG records brain activity at the millisecond level, and therefore is considered to be a strong tool to provide a direct measure of the dynamic interaction between the brain and other stimuli in real-time [203]. It is considered to be a non-invasive method as it can be recorded by electrodes of varying numbers placed on the scalp. However, the level of obtrusiveness can be discussed.

The amplitude of the EEG signals ranges between 10–200 V, with a frequency falling in the range 0.5–40 Hz. There are five frequency bands:


It is also important to note that the precise range assigned to these bands can vary across studies [206] but apart from the exact thresholds, the classification stands.

The importance of these bands is that they are associated with particular cognitive processes. For example, Nyhus and Curran [207] states that theta and gamma bands are correlated to memory processes such as retrieval and encoding, Jensen et al. [208] mentions the connection between alpha and gamma waves and visual processing and Doesburg et al. [209] addresses the whole scalp gamma frequency synchronization in association with consciousness. According to Hamid et al. [210], the presence of stress has been considered to be responsible for an increase in the EEG beta band power.

EEG is a spatio-temporal biosignal, meaning the data not only vary with time but also in location. Where the signal originates at or the comparison of activities in different hemispheres can carry valuable information. An example relevant to stress studies is frontal alpha asymmetry (FAA), which is defined as "the difference between right and left alpha activity over frontal regions of the brain" [211]. It is speculated that the greater left frontal activity is associated with reacting to positive stimuli while the greater right frontal activity is associated with the tendency to withdraw from responding to negative stimuli. The extent of asymmetry has been suggested to vary under conditions of chronic stress and alpha asymmetry is suggested as a potential biomarker for stress classification [212].

Another use of spatial EEG data is when mental workload or performance are presented as evaluation parameters. In task-based experiments, the researchers look at the frontal lobe since as mental demand increases, so does the theta band activity [105]. By recording and comparing the rise in activity, it is possible to quantify the physiological impact of the stressor on the human body. The use of EEG signal to identify attention and distraction is relatively common, as seen in distraction studies in the automotive industry [89,98].

There are methods reported quantifying human acute stress in response to induced stressors (such as impromptu speech, examination, mental task, public speaking, and the cold pressor test) using EEG signal recordings, but the literature lacks a classification of long-term stress using EEG [212]. Even more so, very few of the acute stressors in the literature are related to IEQ parameters (i.e., [2]).

However, studies exist outside the realm of building physics, as [213] study the EEG signals of cold and warm sensation on skin, using 15 healthy subjects. During the experiments, subjects were partially exposed to a temperature range of 14 to 48 ◦C, with different intervals and intensities. While the temperatures, particularly in the warmer range, are not usual from an IEQ point of view, the study provides valuable insight as to how brain activities react to thermal stimuli.

Finally, it is worth mentioning that EEGs, similar to many other biosensing methods, are prone to noise and artefacts that may obscure the data and their interpretation. These artefacts may be caused by the test subjects' movement (eye movements, shivering, coughing, hiccupping, breathing) or due to the faults in the equipment (electrode popping, cable movements, electrical or electromagnetic interferences) [214].

• Heart: Electrocardiogram (ECG) and Heart Rate Variability (HRV)

An electrocardiogram (ECG) measures the electrical manifestation of the ionic potential of the heart, via numerous electrodes placed on the body surface near to the particular organ (e.g., chest, hands, and legs) [215].

Each cardiac cycle in the ECG is characterized by successive waveforms, known as P wave, QRS complex (including Q, R and S waves occurring in rapid succession) and T wave (Figure 5). These waveforms represent the depolarization and repolarization activities in the heart's cells of atrium and ventricle [216].

**Figure 5.** A typical electrocardiogram (ECG) signal that includes three heartbeats and the information lying in the P, Q, R, S, and T waves [after 215].

Studies using the heart as a biosignal data source either use heart rate, or heart rate variability. Heart rate is the measurement of heart beats per minute (bpm), while heart rate variability is the fluctuation of the length of consecutive heartbeat intervals [217], also known as R-R intervals [218] or N-N intervals.

These intervals are not periodic; however, the variation is not random either. The oscillations of a healthy heart are deemed complex and dynamically changing, depending on the "extrinsic protocols imposed on the heart" [219], since the cardiovascular system plays a vital role in reacting to stressors and maintaining the state of homeostasis [220]. Due to this reason, HRV is not only a factor of the heart, but also a rich data source with information on actions of the nervous system in response to stress factors [221].

Accordingly, the use of HRV signals plays a vital role in stress assessment research [183], with several studies being published not only generally in stress literature, but also in the field of building engineering, since the actions of the nervous system in the control of cardiac activities are sensitive to changes in temperature.

As Yao et al. [201] explains, environmental temperature can have an impact on the vagal and sympathetic nerve activity; as under thermally uncomfortable situations, sympathetic activity prevails, while in comfortable situations, the vagal activity overcomes. In humans, the vagus nerve is for the protection of the body, while the sympathetic nerve is triggered for stress response, including thermal stress. Thermoregulation of the human body as a system is controlled by the sympathetic nerve.

The way activities of vagal or sympathetic nerves are presented in HRV are via high or low frequency components. The vagus nerve is thought to excite the high frequency (HF) component of HRV while the low frequency (LF) is thought to be of both. As a result, any physiological reaction concerning thermoregulation triggers the use of a sympathetic nerve, resulting in a higher LF power and lower HF power, and the ratio of LF/HF increases. For this reason, studying the LF/HF ratio data becomes a reliable parameter to understand thermal sensation and physiological comfort in humans [222].

Studies conducted by [201,222–224] have all come up with similar results, contributing to the findings that "higher LF/HF yields unpleasant thermal sensation and discomfort". In Nkurikiyeyezu et al. [225], the researchers had 17 subjects doing light office work under cold, neutral, and hot environments while collecting data on statistical, spectral, and nonlinear HRV indices. With help from machine learning classification algorithms, the study concludes that it is possible to predict people's thermal states in a reliable manner, which was found to be up to 93.7% accuracy. There are no real-world studies on IEQ parameters so far; however, experiments focusing on mental stress using HRV data exist with varying success rates [226,227].

• Skin: Skin Temperature (SKT), Thermal Infrared Imaging (TII), Electrodermal activity (EDA)/Galvanic Skin Response (GSR)

Human skin is a significant source of data when it comes to understanding physiological reactions of the body in relation to its environment. It is the primary sensory organ and anatomical interface the humans have with their immediate environments. The sensations on the skin reflect both the biological processes happening intrinsically, or how the body is impacted by the outer stimuli [228]. To illustrate the skin's adaptability, the operational temperature ranges can be compared. As previously mentioned, the core of the human body requires a rather narrow temperature range, between ~36 to ~40 ◦C, with normal core temperature being ~37 ◦C, and it has been established that the skin temperature can vary between ~15 to ~42 ◦C without sensation of pain [228].

Similar to HRV, skin is part of the thermoregulation system, which is part of the autonomic nervous system (ANS). Therefore, skin-originating biosignals (skin temperature, electrodermal activity or galvanic skin response, photoplethysmograms) can provide variable insights on ANS activity [229].

Amongst these biosignals, skin temperature has been the one investigated most extensively, as it provides a direct understanding in human thermal sensation and comfort estimation, while the sensors are quite inexpensive [230]. Use of contact thermometers to acquire skin temperature has proven successful as seen in [231–235], constant thermocouples in different forms and both in real-world and laboratory settings seem to function and the results seem to be good indicators in predicting thermal sensation. However, thermocouple sensors require electrodes on the skin, making the data collection process slightly obtrusive [188,230].

As an alternative, sensing skin temperature remotely is a method being utilized by several researchers. Infrared imaging technologies are widely available today, and medical infrared has utilized the heat signature of skin to map skin temperature since the 1960s [171]. Current technology presents low-cost no-contact and non-intrusive sensors; however, there are reported application problems: Li et al. [230] states infrared thermometers have a handicap of having such narrow field of views, meaning the thermometers need to be placed very close (few centimeters) to the test subjects. Alternatively, the use of thermographic (thermal) cameras can be placed away from test subjects, while in that case, data accuracy significantly drops in comparison to thermocouple sensors or even infrared thermometers. Nevertheless, with correct data analysis methods, improved data still seem to have been proven to be a robust alternative for real-time thermal comfort detection and even prediction.

One final biosignal of interest uses the skin's thermoregulatory response as the data source. Electrodermal activity (EDA), also known as galvanic skin response (GSR), measures the changes in electrical properties of the skin caused by eccrine sweat gland activity, which is a key factor in the thermoregulatory process [236]. The idea dates back to the very beginning of the 20th century to Carl Jung, who for the first time mentioned electrodermal activity in connection to emotions in a psychoanalysis book [237].

Several studies successfully link EDA data to stress detection and quantification [238–240]. Recently, the use of wearables is becoming more reliable in the field, leading the way to long-term EDA monitoring as well [240].

• Chemical: Cortisol

There are two main pathways in the human body to relay the stress from the brain to the body. The first way is via the sympathetic nervous system, through which cardiovascular and skin-related biosignals were used as stress signifiers. The second way is through the hypothalamic–pituitary–adrenal (HPA) axis, the hormonal route. Stress triggers the HPA axis, a neuroendocrine system that regulates central and peripheral homeostatic adaptive responses to stress [126,129,241,242].

Biochemical samples, primarily urine, saliva, and blood samples, and particularly analysis of hormones such as cortisol and alpha-amylase, are amongst the primary measures used to identify the impacts of stress on the body in conventional psychology methods and long-term stress studies [183,212,243].

While these analyses are deemed to provide high accuracy in stress detection, data collection can pose difficulties in regard to the intrusiveness of obtaining the samples, or the time interval of the data acquisition. Additionally, Lee et al. [244] stress the importance of the timing of the sampling, as cortisol levels in blood also vary diurnally, increasing during early morning and decreasing towards the night. Another important aspect is that while the biochemical signals from fluid analysis may correlate with acute stress, to understand the long-term chronic stress impacts some researchers have suggested looking at extraction of cortisol from hair fibers [245]. In order to overcome the limitations of the intrusive nature of data collection, the use of wearables in the forms of sweat patches, wristbands, or epidermal sensors is becoming more available. As Seshadri et al. [190] remark, while this enables researchers to collect biologic data continuously and is truly non-intrusive, the majority of these devices are only available as market products and have not been clinically tested and validated yet.

#### **4. Discussion**

#### *4.1. Summary and Research Gap*

Our understanding of indoor environmental quality is as good as the methods we use to assess it. A brief review of the state-of-the-art has highlighted how little consensus there is concerning the actual measures used, both between radically different fields and within fields of research. Recurrent criticisms are mostly focused on two main aspects: (i) the effect against which quality (or lack thereof) is measured; (ii) the reliability of the collected data. In the first case, human health and wellbeing are commonly implied to be the end-objective, whether the effects measured vary broadly and include self-selection, performance-oriented or physiological metrics. The use of different measures raises on one hand questions concerning their relevance and appropriateness in reflecting health parameters; on the other, it makes comparisons between studies very difficult. In the second case, the importance of taking into account psychological and individual factors in the equation is on one hand underlined by many researchers as not being adequately considered; on the other hand, these components are widely accepted as having a strong potential for data bias.

This gap in terms of measuring efficiency substantially affects the mental framework of all professionals dealing with the built environment, from designers to policymakers, with end effects not only on human health and wellbeing, but also on the overall energetic and environmental building performance.

Within this context, recent advancements in sensing technology and time-sensitive biomedical data acquisition have opened up new possibilities to use biomedical signals as an increasingly reliable objective methodology to measure environmental stressors. The promise of parsing *physiologically* relevant data is so that the impacts of additional variables, such as adaptive behavior, tolerance, psychology, exposure time, can be studied against a constant, in a situation where, as previously explained, almost all other parameters vary. The use of these methods is however relatively new to the field of building and climate engineering, where the established scientific practice largely uses self-selection or performance-based assessments (questionnaires and interviews) to measure the quality range of measured indoor conditions.

The main hypothesis supported by this paper is that introducing data collection based on biomedical signals to measure human stress in indoor environments has the potential to give new insights, more data reliability and enable comparisons between studies in the field of building physics. Hence, the review aims to contribute with a systematic and comprehensive insight on the current capacity to detect stress from the built environment using biosignal measures, in order to set a foundation for the development of this method of data collection in the field of building physics.

#### *4.2. Approach*

This paper reviews the state-of-the-art knowledge concerning (i) indoor environmental parameters affecting humans in terms of IEQ and (ii) human biosignals that respond to the environmental IEQ stressors and the physiological response they activate, and then maps the interaction and interdependence between the two. More than 240 scientific publications were analyzed, published between 1906 and 2021, dealing with topics ranging from built indoor environment, medicine and neuroscience, biomedical engineering, sensing technology, psychology, ecology, and economy.

#### *4.3. Main Findings*

Indoor environmental parameters commonly used in building engineering and indoor climate research were reviewed. Indoor environmental quality (IEQ) is determined by assessing air quality, thermal, visual, and acoustic parameters. Although the use of these parameters is well-established throughout the reviewed literature, this brief review was necessary in order to establish a common ground between current practice and the introduction of biosignal measures. This will allow bridging the new methods introduced and the state-of-art knowledge on the features of the indoor environment affecting humans, but also to set a basis for comparison with the existing research in the field.

Moreover, discussing IEQ from a health-oriented rather than a comfort-oriented perspective, relevant sub-parameters were identified, which are, in the current state of research, not taken into consideration. This is the case of visual IEQ, where the range of the natural

light spectrum should be taken into consideration also for its health-related properties (as to trigger the production of vitamin D) rather than only for its optical properties.

Parallelly, human biosignals were reviewed. Biosignal definitions and classifications were introduced in order to provide a systematic framework for professionals dealing with the built environment. The physical nature as well as the level of invasiveness of the biosignals are discussed. The physical nature of biosignals directly relates to the physical IEQ aspects of the indoor environments, while the invasiveness of the technology enables an easy and unbiased data collection that does not interfere with the study subjects' behavior patterns or awareness of being observed.

Successively, the review reported on the main limitations connected to the use of biosignal measures, being mainly of technical nature (noise, wearability, and accuracy), country-specific legal issues involved with data collection and storage and the use in some contexts of self-reporting as validation. The necessity of relying on multimodal data acquisition is also highlighted as a possible limit; however, if systematically integrated in the research method, it remains a valid method for cross validating the acquired data. The possibility of performing this kind of analysis can therefore also be seen as a strength, yielding potentially high quality and complete sets of data.

Finally, the review reports on the specific biosignals that relate to the IEQ parameters. These are classified into four categories depending on the biophysical aspect measured: brain, heart, skin and chemical.

#### *4.4. Limitations*

The introduction of biosignals as measures of indoor environmental quality still appears to have a number of limitations, in spite of the potential benefits highlighted by this review.

Defining the metrics of environmental quality remains a complex task due to a number of intrinsic complexities:


Additionally, a lack of knowledge and training from the side of building professionals in terms of the medical and technical background necessary to correctly use some biosensor measures, to collect and to interpret the data is a main limit to the effective use of these technologies. This problem is especially evident with some biosignals, such as the EEG, where possibly specialists from other fields need to be involved. Although the interdisciplinarity of these types of studies has the potential to contribute with innovative practices and relevant findings, it poses important limits in terms of time and resources that can discourage the use of these methods in regular ordinary research.

Finally, the study has addressed "indoor environments" as a whole, not distinguishing further between the type of environments, in the aim of presenting the method as an overarching tool that can be successively adapted to more specific case-studies. From the analysis of the reviewed literature, a great majority of the research has up to today focused on office spaces, a few on housing indoors and another few on healthcare contexts.

Intrinsic limitations of the review itself, in spite of the high interdisciplinarity of the sources reviewed, might come from the authors having similar cultural and professional backgrounds. Additionally, the studies reviewed were exclusively written in the English language, with the potential exclusion of relevant sources published in other languages.

#### *4.5. Future Directions*

This review aims to set a foundation for the introduction of biosignals for the assessment of environmental parameters in the built environment, but is in that sense only a first step within a long process. Follow-up studies to effectively establish the biosignal data collection methods into current practice include:


#### *4.6. Conclusions*

This review has highlighted how the relationship between daily human indoor habitats and human health needs to be researched further. While in current practice, comfort studies prioritize optimizing building energy use and user satisfaction, and studies on productivity bring a more direct economic impact, both constellations forgo the health implications as they are not visible at the same timescales. In other words, a balance needs to be found between the right dose of a higher overall human resilience and building resilience [246].

The use of biosignal measures to detect environment-related stress conditions is put forward as a promising and reliable method to effectively focus on short and long-term human health aspects. The review is a first stage within a longer process of translation of a method that is already established in other fields of study using stress research in the context of the built environment.

**Author Contributions:** Conceptualization, S.G.L.P., B.K., S.C.K. and T.A.; methodology, S.G.L.P. and B.K.; formal analysis, S.G.L.P., B.K. and S.C.K.; investigation, S.G.L.P., B.K. and S.C.K.; resources, S.G.L.P., B.K. and T.A.; data curation, B.K. and S.C.K.; writing—original draft preparation, S.G.L.P., B.K. and S.C.K.; writing—review and editing, S.G.L.P., B.K. and S.C.K.; visualization, B.K.; supervision, S.G.L.P., B.K. and T.A.; project administration, S.G.L.P., B.K. and T.A.; funding acquisition, S.G.L.P., B.K. and T.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded through contributions of the ALEXANDER VON HUMBOLDT STIFTUNG foundation, Bonn, Germany, through a research grant to Sandra G. L. Persiani, grant number ITA 1211263 HFST-P and of the TUM SEED FUND RESEARCH through a research grant to Bilge Kobas.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

## **Abbreviations**


#### **Appendix A**

**Table A1.** Sources based on their main research fields.



**Table A2.** Sources about comfort, wellbeing, stress, and other definitions, based on their main research fields.


**Table A3.** Sources about human aspects, categorized based on their focus.



**Table A4.** Sources about indoor environmental quality (IEQ), based on their publication types.

**Table A5.** Sources about IEQ, based on their main research fields.


**Table A6.** Focus points of the publications on IEQ.



**Table A7.** Types of IEQ parameters in publications on IEQ.

**Table A8.** Sources about physiologic signatures, based on their publication types.


**Table A9.** Sources about physiologic signatures, based on their main research fields.


#### **References**

