Vijayakumar Nanjappan, … Ka L. Man, in Big Data Analytics for Sensor-Network Collected Intelligence, 2017
3.2.10 Wearable devices
Wearable devices or gadgets are electronic devices worn by consumers ubiquitously and continually to capture or track biometric information related to health or fitness. Wearable devices are new manifestations of accessories that people wear, such as Apple’s Watch or Samsung’s Gear Watch or more dedicated tools like the Fitbit One wireless activity and sleep tracker and monitor. Wearable devices with biometric tracking capabilities represent one of the most important sources of data generation. They will continuously and uninterruptedly record data of different types and from a variety of environments.
As data increase in variety and volume in parallel to the need to support greater velocity in their generation and processing, it is important to have a way to organize them. Organization and management of data will therefore be explored next.
Rajesh Singh, … Mamta Mittal, in Wearable and Implantable Medical Devices, 2020
1.2 Intelligent wearable device description
The IWD pedometer aims at efficient calculation of steps and calories from accelerometer signals after filtering out noise. It utilizes the low-power mode of the microcontroller and sensors to achieve a better power-efficient pedometer algorithm. Fig. 1.1 shows the generalized block diagram of the IWD network. Multiple IWD devices are proposed, which may be connected to a common server, to monitor physical activity at a centralized place. The controller consumes less power in idle state.
Figure 1.1. Generalized block diagram of intelligent wearable device (IWD) network.
Fig. 1.2 shows the block diagram of IWD, including the microcontroller, accelerometer, and battery. Fig. 1.3A shows the block diagram of the IWD smart system, which is comprised of Arduino Nano, NodeMCU (WiFi device), Bluetooth and LCD20*4, and battery as power supply. NodeMCU communicates the data to the cloud server as well as to the app.
Figure 1.2. Block diagram of the intelligent wearable device.
Figure 1.3. (A) Block diagram of the system. (B) Hardware schematic of the intelligent wearable device (IWD).
Fig. 1.3B shows the schematic diagram of the proposed IWD algorithm.
Shulong Ye, … Peng Yu, in Titanium for Consumer Applications, 2019
2.3 Wearable devices
Wearable devices are products controlled by electronic components and software that can be incorporated into clothing or worn on the body like accessories. Nowadays, a variety of wearable devices, such as smart glasses and smartwatches, have been invented. The sales of these consumer electronics have been growing steadily. In the fabrication of wearable products, the utilization of titanium can increase the accessional values of these devices. Materials with superior corrosion resistance and good biocompatibility are considered suitable for wearable devices frequently in contact with human skin. Titanium products are resistant to corrosion by perspiration, which makes titanium an ideal material for wearable devices used in the sports. Moreover, the high specific strength of titanium makes the frame of the devices strong enough to protect the electronic components inside and significantly decreases the overall weight of the devices.
Fig. 5 demonstrates a titanium glasses arm for augmented reality smart glasses fabricated by metal injection molding (MIM) [16]. MIM combines the merits of powder metallurgy and plastic injection molding, and offers design freedom in shape complexity and is suited for mass production [17, 18]. The glasses arm can be made of CP Ti or Ti-6Al-4V alloy. Fig. 5A shows the product design of the augmented reality smart glasses whereas Fig. 5B is a photo of the MIM-fabricated titanium glasses arms. The glasses arms are anodized to obtain different colors, during which titanium reacts with oxygen to form an oxide layer. The thickness of the oxide layer can be adjusted by heat or electrochemical treatment. When the thickness of oxide varies, the reflected light waves vary resulting in different colors of the part. It is worth mentioning that the titanium glasses arm is more than 170 mm long but weighs only 6.8 g, featuring an extremely complex curved surface and a thin wall structure with smallest wall thickness ~ 0.5 mm. Fig. 5C and D shows SEM micrographs revealing the complex details in the glasses arm. Therefore, it is both difficult and expensive by machining and other processing methods, such as forging or casting, to mass produce the component. In fact, it is challenging to the MIM process as well. These MIM-fabricated glasses arms can achieve tolerances consistently controlled within ± 0.5% of the nominal dimensions, as well as good densities and mechanical properties comparable to those of the cast counterparts.
Fig. 5. (A) Product design of smart glasses, (B) a photo of the Ti glasses arms after coloring, and (C) and (D) SEM micrographs showing delicate structures of the MIMed Ti glasses arm.
(Courtesy of ElementPlus, Shenzhen, China.)
There are other examples of titanium wearable parts made by MIM. Sintered to a high density, the watch frame weighs ~ 6 g and has a curved surface and internal complex structures. All the complex structures are net-shape-formed without machining. After being carefully designed to prevent dimensional distortion during sintering, tolerances controlled within ± 0.5% of the nominal dimensions can be consistently achieved.
Jingjing Shi, Jianqing Wang, in Wearable Technology in Medicine and Health Care, 2018
11.10 EMC Test Method for a Wearable Device
Wearable devices operated in a body-centric network system require EMC testing to ensure that they can be used for vital signal collection and transmission without failing or causing other devices to fail. So in such a humanized environment, existence of human body cannot be disregarded for EMC test. However, basically in a practical EMC experiment, realistic human body is not permitted, so that the generation of a series of pseudo-signals which can simulate realistic vital signals of human body is necessary, particular in an EMC immunity testing.
Fig. 11.12 shows the composition of our developed pseudo-vital signal generator. The vital signal such as ECG or EMG signal, gathered as a series of digital data with a sampling frequency of 2 kHz and quantization level of 8 bits, is stored as preliminary preparation in a PC for its reproduction. After conversion with optimal amplitude, this digitized vital signal will be sent to the control circuit through serial port, and then the digital-to-analog (DA) converter to reconstruct the analog vital signal with a shaping filter. In this way, it is possible to simultaneously output the reproduced vital signals at a maximum of four channels. As a verification example, the EMG signal reproduced by the pseudo-vital signal generator is shown in Fig. 11.13, to compare with the realistic EMG signal. A perfect match can be seen obviously there, and the correlation coefficient between the reproduced pseudo-vital signal and realistic signal was found to be as high as 0.99. Such a well-reproduced pseudo-EMG signal can provide a good insight into the immunity testing for application of myoelectric-controlled arm prosthesis. In this case, the pseudo-EMG signal can be learned and translated into information by myoelectric prosthetic arm, and then the electric motors can use the translated information to control the artificial arm movements as one expected. Of course, the possibility of using a pseudo-EMG signal in an immunity testing for myoelectric-controlled arm prosthesis has been confirmed experimentally.
Figure 11.12. Composition of pseudo-biological signal generator for EMC testing.
Figure 11.13. Comparison of the myoelectric signal reproduced by the pseudo-biological signal generator and the realistic one.
For the wearable devices attached on human body, we thus can provide a basic immunity test method as depicted in Fig. 11.14, an illustration of immunity testing system using pseudo-vital signal generator and biological tissue-equivalent phantom. A solid biological tissue-equivalent phantom is used as the substitute of realistic human body. In order to make the biological tissue-equivalent phantom act as human, the pseudo-vital signal generator is embedded into the phantom by means of connecting with two output signal electrodes inside the phantom. In this way, immunity experiments for the wearable devices under test can be conducted to verify whether or not they cause malfunction for the detection and transmission of pseudo-biological signal. Therefore, instead of realistic human body, our suggested testing method by using a biological tissue-equivalent phantom incorporated with a pseudo-vital signal generator can provide a good solution to deal with these kinds of wearable devices for engineers in the corresponding EMC tests. Since the international standards on EMI and EMC test are still under way, this work also contributes to providing suggestions and promoting international standardization process for wearable medical devices.
Figure 11.14. Mechanism of EMC testing for a wearable device based on pseudo-vital signal generator and biological tissue-equivalent phantom.
Xiaoming Tao, in Wearable Electronics and Photonics, 2005
Output interface
Wearable devices have output interfaces by which information is presented to the wearer. Vibration (tactile) interfaces have been used. An example of this is the vibration function in mobile phones, by which the user is silently alerted to an incoming call. Many portable devices use audio interfaces. In both cases, the amount of information given is quite small. Voice synthesis (the opposite of voice recognition) via earphones is an alternative, as the wearer does not need to decode the message and can understand it directly. A third category of output interface is the visual interface. These include, for instance, seven-segment or dot matrix displays, liquid crystal displays (LCDs), organic and polymeric light-emitting diodes (OLEDs and PLEDs), and fibre optic displays (FODs). The displays may take two forms: wearable flat panel displays or head-mounted displays.
The main display technology used in portable electronics today is the LCD screen. It is neither flexible nor lightweight. Moreover, it can be bulky and its angle visibility is poor. Holographic polymer dispersed liquid crystals (HPDLCs) are still in their infancy; however, they may offer better performance in terms of flexibility. Polymer light-emitting diodes (PLEDs) are very promising candidates for future wearables, as they have high contrast, a high level of brightness, require much less power and are flexible. Flexible displays based on polymeric fibre optics are also being investigated by a number of researchers.
Electroactive polymer actuators take the form of fibres, yarns and structures based on thin film. They are used as artificial muscles for robotics. According to their actuating mechanisms, they can be broadly divided into two groups: electronic and ionic. The electronic polymers include electrostrictive, electrostatic, piezoelectric and ferroelectric polymers. They can hold induced displacement when a DC (direct current) voltage is applied and have a high level of energy density in air. However, a high activation field greater than 150 V μm−1 is required. Ionic polymeric materials include polymer metal composites, conducting polymers and polymer–carbon–nanotube composites. They normally perform actuation in a solution and have a low activation voltage of 1–5 V μm−1. All of these actuators have limitations for use in wearable devices. A promising new technology is based on the dielectric elastomer, which is activated with low voltage in the air and is very robust and flexible. Books by Tao (2001) and Bar-Cohen (2001) provide very comprehensive accounts of dialectric elastomers.
Adam Bohr, … Henrik Jensen, in Microfluidics for Pharmaceutical Applications, 2019
6.1 Increased Interest in Wearables
Wearable devices are becoming increasingly popular for health care purposes. The growth of industries, such as remote patient monitoring devices and home health care, is expected to influence demand for wearable medical devices over the course of the next years. The explosion in wearable technology has mainly been seen in relation to health care monitoring, but wearable technology is also relevant for diagnostic and therapeutic applications. The wealth of continuous and real-time data collection can help classify patient populations resulting in higher efficacy of treatment and increased safety. Social media is also having an impact on the use of wearables and health care monitoring encouraging its expansion into the everyday life for both the healthy and the ill, helping to understand the needs of the users [122]. The high demand for wearable medical devices and remote patient monitoring has fueled numerous possibilities, with an estimated 113 million units of wearable devices shipped in 2017 alone, with a market value of approximately USD 20 billion [123,124].
Before describing the development and application of smart wearable systems, it is important to clarify what the term, wearable system, actually covers. In the broad sense, wearable systems are systems that have one of the following features: wearable, portable, implantable, and ingestible. They can comprise a patch, sensor, actuator, fabric, textile, fiber, clothing power supplies, wireless communication, multimedia devices, software, etc. They can be worn as an accessory or embedded as part of clothing, carried around and used as needed, implanted in the body or ingested as a pill [125]. They can be used at home, outdoors, work, or clinical setting and can be connected with mobile or stationary equipment.
So far, smartwatches and wristbands have been the focal points of mass-market consumer product development within wearables, which has led to an abundant availability of such devices in the past couple of years. The main players here are companies including Google, Samsung, Nike, Apple, and Fitbit. Although these devices are mainly focused on fitness applications and other activities, there is an increasing demand for wireless monitoring devices, and an increasing incidence of lifestyle-related diseases requiring routine vital statistics and analysis is driving the interest toward health care-related application of wearables. Research within smart wearable systems has concentrated on smart devices and intelligent environments, which displace the interface for computational surroundings to our body, clothes, and portable accessories. Wearables for health care purposes are likely to follow the same trends where small miniaturized devices with a bare minimum functionality can be used together with computation in the environment, for instance, a smartphone [124]. Here, adaptability can be regarded as essential feature of wearable systems for health care purposes, and these should be compatible with smartphones and other devices. Intercompatibility among wearables and standardization of the obtained health care data will most likely also be important features for bundling and correlating all the data acquired in a more holistic health care monitoring approach (Fig. 15.7).
Fig. 15.7. Wearable microfluidic devices. Miniaturized microfluidic systems can be integrated in (A) dermal, intradermal, circulatory, and intracranial devices for diagnostic, medical, and drug delivery purposes. Monitoring and device control can be exerted through user-controlled accessories (smartphone, smartwatch, etc.) or remotely by health and law enforcement organizations. Examples of medical microfluidic devices are (B1) wearable mixers for self-administration of drugs at flexible dosages/composition; these can also be integrated in micronized delivery devices, that is, (B2) oral devices for intestinal injection of instantaneously mixed stabilized proteins with excipients. Microfluidic surface properties can be also exploited in (C1) wearable dialysis devices or electrolyte exchange devices. Microfluidic channels with (C2) surface-associated antibody and enzyme-based microfluidic systems could provide a wearable system to reduce the spreading of circulating cancer cells or deplete tumor growth.
Avishek Choudhuri, … Shankey Garg, in Internet of Things in Biomedical Engineering, 2019
6.3 Applications and Technologies
Detecting and managing chronic diseases and long-term conditions is a major healthcare issue. A rapidly growing technology application in healthcare is the use of individual medical devices to monitor individual patients at home, often called ubiquitous healthcare. The data collected from the patient is sent to a dedicated server, where it can be accessed by medical personnel, caregivers, or other appropriate persons, analyzed, and appropriate actions taken. The concept of applying the IoT to ubiquitous healthcare can enhance monitoring abilities by displaying the data in a meaningful format and allowing an immediate and automatic exchange of data with other network devices and systems (smartphones, portals, servers) so that diagnoses and other responses can be carried out [28].
Healthcare is a very data-intensive field. Often in medical diagnosis, reaching a correct diagnosis requires information of prior health conditions and findings. Metabolic disorder alludes to a condition in which the patient exhibits three or more of the following conditions: high glucose level, hypertension, obesity, high cholesterol, and high triglycerides [28, 29]. Metabolic disorder can cause various complications, for example, cardiovascular disease or stroke [29]. As another example, blood pressure is typically considered to be at Stage I hypertension when the systolic measurement is more than 140 mmHg and diastolic is more than 90 mmHg [28, 29], but patients with diabetes or kidney disease are diagnosed with Stage I hypertension if the systolic is over 130 mmHg or diastolic is 80 mmHg. Hence, diagnosing hypertension requires the gathering of complete medical information (diastolic and systolic blood pressure), considering the age, sex and prior health records of the patient, for example, weight, diabetes or other chronic diseases. For diabetes type 2 to be diagnosed, for instance, we need to consider risk factors, for example, weight, alcohol consumption, age and sex, infections, or hereditary illnesses [30].
Improving human healthcare systems and infrastructure worldwide is one of the most important and challenging applications of the IoT. Many issues are involved, but the main goals are to provide better quality care to patients and lower healthcare costs; helping to alleviate nursing and other staff shortages is another important parameter. According to Ref. [31], in fact, patient observing, assessment, administration, and many other healthcare tasks are regularly manually handled by nursing staffs, which often experience shortages, leading to productivity bottlenecks and hence to possible errors, with dangerous consequences. Ongoing technology advances in the structure of the IoT are prodding the advancement of smart systems for healthcare and biomedical- related procedures [32]. Automated remote monitoring and tracking of patients and biomedical devices in hospitals, telemedicine services as already discussed, and monitoring of elderly or frail patients at home, which enables recognition of clinical deterioration at early stages, are just a few of the important applications. Among others, ultra-high-frequency (UHF) radio frequency identification (RFID), wireless sensor networks (WSNs), and smart mobile devices are three of the most encouraging technology advancements empowering the execution of these smart healthcare systems.
RFID is a low-power, low-cost device comprising an embedded antenna and an IC chip along with an identification (ID) code, called tags, that can transfer data when powered by the electromagnetic field generated by a reader. There are several types of tags. Passive RFID tags do not require a power source, and their lifespan can thus be measured in decades, but they have a short range. With battery-assisted passive (BAP) tags, the range is longer, up to 100 m under ideal conditions, but they have a shorter lifespan due to the battery. In addition, new advances such as sensing and computation add value to this technology. RFID has a wide range of applications, healthcare being a major one [33–35]. Active research is ongoing into many RFID uses [36–38], especially regarding IoT-related applications. RFID-based monitoring in healthcare empowers zero-control, minimal effort, and simple-to-execute observing and transmitting of the physiological conditions of patients. The main downside of RFID labels is the short distance over which they can operate, i.e., only 15–25 m. Plainly, this factor restrains the usefulness of UHF RFID devices to object/patient identification and monitoring within small areas [31, 33–46].
WSNs are spatially distributed systems of autonomous sensors that can record and transmit physical or environmental data. The individual sensor nodes are coordinated in a multijump design to provide screening and controlling functions for many applications, including mechanical, military, home, vehicle, and healthcare situations. Currently, most WSN nodes are battery-powered transducers incorporating simple/computerized sensors and an IEEE 802.15.4 radio empowering about 100 m range (one jump). Contrasted with the coordinating, detecting, and processing capacities of UHF RFID labels, WSN nodes expend more power and have a shorter lifetime [39–41].
In healthcare applications, the combination of RFID and WSN may provide more expanded applications and broaden the scope of uses [42, 43]. To our knowledge at the time of this writing, only a few research studies have been carried out using the combination of UHF RFID and WSN advances in healthcare applications. Based on this idea, IoT devices would remotely open through the Internet, thus permitting the improvement of applications and allowing further innovations for the programmed monitoring and tracking of patients, staff, and biomedical devices within hospitals, nursing homes, and other healthcare facilities. Such applications can be made possible through the interaction of various technologies including the combination of RFID and WSNs, mobile cellphones, the emerging standard 6LoWPAN, and CoAP. In particular, structured SHS can continuously gather both environmental and patient physiological data through an ultralow-control hybrid sensor network (HSN) of 6LoWPAN hubs incorporating UHF RFID Class-1 Generation-2 (called Gen2 from this point forward) functionalities.
Such a system incorporates an enlarged RFID Gen2 tag with the end goal of storing sensor information and patient data. Along these lines, physiological patient data can be collected using RFID Gen2 readers distributed by the clinic and conveyed to a control center where an application allows the data to be readable by both local and remote clients by means of a REpresentational State Transfer (REST) Web architecture. Amid typical tasks for such a system, no WSN-based transmitting needs to be performed, consequently decreasing hub control utilization. The modeled device is capable of managing emergency situations in a timely and reliable manner. In fact, for such a situation, transmission based on WSN would be enacted in order to alert the nurse or other healthcare personnel by means of Push Notifications on a portable application. Specialists could likewise connect a smartphone to a convenient UHF RFID reader and utilize a similar mobile application to connect with patient hubs during daily routine health examinations [44].
As highlighted in Ref. [31], currently patient care, observation, managing, and supervising are physically implemented by a nurse. This can prove to be a barrier and bottleneck and even a factor in devastating medical practice errors. New IoT advances are impelling improvements, namely smart systems and devices to enhance healthcare and biomedical-related procedures [33–35, 45].
The treatment of chronic medical illnesses like diabetes, arthritis, and chronic obstructive pulmonary disease (COPD) requires constantly increasing amounts of healthcare services along with the associated rising costs [47]. In addition, it is a pattern that elderly individuals are living longer and staying in their homes in greater numbers. Novel E-health IoT applications empower continuous monitoring of patients with chronic diseases and the frail elderly, especially those living alone. Smart applications of sensor innovations are central to the new IoT applications. Utilization of these advances can decrease the response time during healthcare emergencies and improve self-care outcomes. Use of these sorts of IoT applications incorporates numerous security system components, complex and effective programming, and is liable to high security protections [48].
6.3.1 Biomedical Signal Compression
Wearable devices and smartphones tend to have very tight limitations on storage, network connectivity, computational power, and battery usage, and thus need efficient storage and communication, creating a challenge for IoT applications in healthcare. More traditional encoding schemes can incur high compression costs in some situations. An innovative low-cost simple data compression scheme for m-healthcare monitoring devices is introduced in Ref. [49]. Using the periodicity found in many medical signals, such as ECG, this method is based on MPEG compression (e.g., video compression) but is especially suited for biomedical and healthcare monitoring. Several studies have focused exclusively on ECG signals, incorporating techniques such as discrete wavelet transform (DWT) [50], discrete cosine transform (DCT) [51], and discrete Fourier transform (DFT) [52]. These methods are not typically suitable for wearables or implanted medical devices because of their high computation demand. Most wearable medical devices have restricted computational capacities and low throughput.
However, Ref. [53] discusses a data compression technique for portable ECG Holter monitors, which are usually worn by patients while ambulatory for up to 2 weeks, using quad level vector (QLV) and the data compression algorithm Huffman coding. Like Ref. [53], researchers in Ref. [54] introduced a compression strategy that incorporates multilevel vector (MLV) compression, an integer linear programming (ILP) plan, and Huffman coding. The ILP plan speaks to a base set-covering issue that picks certain flag tests to be incorporated into the compacted yield, while the others are simply disposed of. The ILP issue, however, is computationally intense and inapplicable for wearable gadgets, and also creates compression losses. The compression technique published by researchers in Ref. [55] claims 60% reduction in data size to be transmitted by radio, which could increase battery lifetime from 7 to 10 days of portable healthcare devices requiring wireless transmission of data.
As already discussed, many smart healthcare devices are available in the marketplace, with many others soon to appear [56–58] for healthcare and wellness services. In addition to the wearable devices available in the popular and recreational market, researchers are working on applications of IoT for medical uses, that can record and organize clinical data from patients and provide access to this data [59–63]. It’s possible to foresee that in a short time (and already, in some locations and facilities) routine medical examinations will no longer require weeks to get test results, instead making use of wearable sensors that constantly monitor and record clinical parameters and store them into a database over a prescribed period of time. Clinicians will be able to view not only the focused/laboratory reports, but more extensive long-term records that sensors provide. Medical experts using this information can recommend more effective treatments or early interventions, that lead to patients making better lifestyle choices, having a positive impact on human health worldwide. These IoT/sensor innovations have existed for just a few years and so far they have had relatively minimal impact on the current medical industry, which has tended to be a slow adopter of change in the past. This chapter focuses particularly on the medical sector and looks at the many opportunities presented by the technologies discussed and the challenges that must be met before they can be realized and become part of routine healthcare for everyone [64].
J. Decaens, O. Vermeersch, in Smart Textiles and their Applications, 2016
23.1.2 Various levels of integration
The addition of wearable devices should not impede the comfort of the PPE or the mobility of the wearer. Therefore, a high level of integration is expected but is not always met. Indeed, the level of intimacy between the electronic components and the textile substrate can be classified in three categories:
1.
A low level of integration implies that the wearable device is added during the last step of manufacturing – the assembly.
2.
A medium level of integration indicates that the functional components are directly embedded within the fabric.
3.
A high level of integration entails the incorporation of the active elements within the fibres themselves.
Increasing the level of interweaving between the components of the wearable device and the PPE can be achieved by acting backward on the manufacturing process steps. However, the level of difficulty also increases and can raise further challenges, such as durability after washing cycles, since the functional elements are no longer removable.
T.M. Navamani ME, PhD, in Deep Learning and Parallel Computing Environment for Bioengineering Systems, 2019
Mobile Devices
Smartphones and wearable devices which are embedded with sensors play a significant role in health monitoring. By using these devices, direct access to personal analytics of the patients is possible, which can contribute to monitoring their health, facilitating preventive care and also helping in managing ongoing illness [2]. Deep learning plays a crucial role in analyzing these new kinds of data. There exist some challengeable issues like efficient implementation of deep neural architecture design on a mobile device for processing data from sensors, etc. To overcome these challenges, several suggestions were proposed in the literature. Lane and Georgiev [61] have proposed a low power deep neural network, which exploits both CPU and digital signal processor of mobile devices without burdening the hardware. They also proposed another deep architecture DeepX, a software accelerator that can minimize the resource usage, which is the major need of mobile adoption. It also enabled large scale deep learning to execute on mobile devices and outperformed cloud based off-loading solutions [62]. In [63], the authors have incorporated CNNs and RNNs with LSTM to predict frozen gait problems in Parkinson disease patients. These patients will struggle to initiate movements such as walking. A deep learning technique was also used to predict poor or good sleep of persons using actigraphy measurements of the physical activity during their awakening time.
Juan M. Fontana, Edward Sazonov, in Wearable Sensors, 2014
5.2 Wearable Devices for Free-Living Monitoring
A novel wearable device, the Automatic Ingestion Monitor (AIM), has been developed and evaluated for objective monitoring of food intake under free-living conditions (Figure 8) [24]. AIM presented three major benefits over self-reported intake. First, AIM is a wearable device that has the ability to monitor 24 hours of ingestive behavior without relying on self-report or any other actions from subjects. AIM wirelessly integrated three different sensor modalities for an accurate monitoring: a jaw motion sensor to monitor chewing, a proximity sensor to monitor hand-to-mouth gestures, and an accelerometer to monitor body motion. Second, AIM is able to reliably detect food-intake episodes in the presence of real-life artifacts using a robust pattern-recognition methodology for detection of food intake. The detection methodology contains several steps, such as sensor information fusion, feature extraction, and classification. The sensor fusion step removes portions of the signal that cannot be food intake based on statistically derived rules. For example, it is highly uncommon to eat solid foods during moderate to vigorous exercise, or during sleep. Both of these activities (exercise and sleeping) can be reliably detected from the accelerometer signal and corresponding signal intervals not included into further consideration for food-intake detection. The feature extraction step computes a number of time, frequency, and time-frequency domain features from the sensor signals. Food-intake detection is based on an artificial neural network implementing a subject-independent classification model that requires no individual calibration. Third, the AIM device and food-intake detection methodology were validated in an objective study where an average food-intake recognition rate of 89.8% was achieved. Individuals with origins from five different countries and having different lifestyles and ingestive behaviors participated in the validation study. They wore AIM in free living during 24 hours without any restrictions on their eating behavior and activities.
Figure 8. The Automatic Ingestion Monitor (AIM) consisting of four main parts: (a) the jaw motion sensor, (b) the wireless module, (c) the proximity sensor, and (d) the smartphone.
The results of the validation study revealed that AIM can potentially provide an accurate prediction of the food-intake episodes occurring over the course of a day in a free-living population. However, several questions remain to be answered. One question is related to the capability of AIM for detecting liquid intake. In the validation study, the results showed the recognition rate for solid food intake only. Previous studies suggest that certain intake of liquids (such as gulping large quantities of a drink) may be detected through the monitoring of jaw motion [33] while others, such as sipping, may be undetected. Another question is related to the acceptance of the device by subjects. AIM was designed as a pendant device worn on a lanyard around the neck, which intended to satisfy the need for a socially acceptable device; further miniaturization of the device is needed to make it less obtrusive. Finally, although the food-intake detection was performed offline, the ultimate goal of AIM is to perform real-time recognition and characterization of food intake and to deliver feedback about an individual’s intake behavior.
The post Wearable Device appeared first on .::KathaBuzz::..