1 Introduction

When visiting a museum, as traditional introductions and content about general exhibits are limited, visitors often have to interpret the connotations of the exhibits by relying on their knowledge, in order to self-determine the relevance of works. As a result, museum tours tend to be superficial. Although some exhibition sites are equipped with voice navigation systems to provide further navigational services (for instance, visitors can manually enter the voice navigation number of the corresponding exhibit to listen to relevant exhibit introductions), the method of voice navigation can only narrate the historical background of exhibits, and its navigation effect is extremely limited.

In recent years, with the increasing popularity of smart mobile devices, mobile tour navigation-related applications have become increasingly common [1, 21, 25]. For example, the Brighton Museums app from the UK provides information on the collectibles of five museums in Brighton. In addition to providing region maps, historical pictures, featured activities, and the latest news about the museums, it can help tourists in planning their tours [1]. A series of mobile navigation systems for the National Palace Museum in Taiwan have been designed, by which users can learn more detailed information about the exhibits through content arrangement, educational games, and audio-video functions [25]. While the above navigation systems can provide more detailed navigation information to make navigation more efficient, the display functions are still limited. The diversified AR multimedia display functions can address and improve this problem well [21].

With the rising hardware performance and speed of mobile devices, many scholars and internationally famous exhibition sites have actively engaged in the development of mobile navigation systems with AR display functions. For example, Chen et al. applied marker cards to implement a museum 3D display system for cultural relics, which can support rotation, magnification, and other interactive instructions [2]. Wolff et al. cooperated with the Irish Museum of Modern Art to develop a mobile navigation system that can identify QR code cards, which tell the stories of the artworks [24]. The Streetmuseum app, as launched by the Museum of London, uses the built-in GPS positioning function of smart phones to display overlapped historical photos of realistic street views, in the manner of AR, in order to show interesting ancient and modern scenes with interlaced virtual and reality combinations [16].

In 2016, the National Palace Museum of Taiwan developed a mobile navigation app with a markerless AR recognition function called Giuseppe Castiglione, which can offer tourists useful navigation information and interesting interactive functions during their visit, such as taking a photo with an AR character [8]. The National Palace Museum developed another NPM InSight App in 2017, which has an AR treasure hunt game and a mobile phone lens can be used to look for and collect the virtual cultural relics to increase the user’s fun during the visit [12]. The Austrian Museum of Celtic Heritage also developed a markerless AR display system in 2016, which can show 3D virtual character animation, and narrate the backgrounds and stories of the exhibits in the museum [13].

On the other hand, while the above-mentioned mobile navigation systems can provide diversified multimedia navigation information, the users often cannot effectively absorb such a large amount of navigation information; therefore, the addition of a recommendation system to provide practical and useful information could be helpful to the user [20]. A number of scholars have developed practical recommendation systems for mobile devices; for example, Willem et al. implemented a mobile museum tour guide system with a recommendation function for providing personalized tours tailored to the users’ interests and their position inside the museum [9]. Su et al. put forward a mobile navigation system with a group recommendation mechanism, where the best navigation route of the tourist attractions can be recommended by analyzing various users’ evaluations [17]. Shen et al. proposed a mobile navigation system combining GPS positioning and a recommendation mechanism, which can recommend the tourism attractions and foods in this region according to the user’s position, and guide the users to choose a region where the visitor flow is currently less, in order to smoothly achieve the tourism experience [15]. Feng et al. used the principle of User-Based Collaborative Filtering to group users by recording their preference for the historical relics they have already visited, and further recommend historical relics that the users may be interested in according to the users’ preferences in the groups [6].

Moreover, while navigation systems can recommend and provide detailed information to make navigation more efficient, general visitors often must find the exhibits and tour routes by themselves, without knowing their locations in real time, thus, they may not be able to find the location of exhibits in large exhibition places, and will waste time searching for them. However, the real time indoor positioning technology, which has emerged in recent years, could solve this problem well.

In general, the different types of positioning systems can be divided into the two major categories; indoor and outdoor positioning systems. Although outdoor GPS positioning technology has matured and become popular, it is only applicable outdoors and cannot offer an indoor positioning function, while the methods of indoor positioning remain under development. Nowadays, the main methods of indoor positioning include Wi-Fi, infrared ray, ultrasonic, Bluetooth (before 4.0), active RFID (radio-frequency identification), and BLE (Bluetooth low energy) beacons (such as iBeacon); of these methods, BLE beacon positioning technology is a relatively better choice [22].

A number of scholars have recently published related studies and applications of BLE beacon positioning [5, 7, 10, 11, 14, 26]. Martin et al. proved the feasibility of indoor positioning with iBeacon [11]. Rida et al. judged the user’s approximate location in a building through the principle of triangulation location [14]. Fujihar et al. stated that many iBeacons are set in corridors, and through the iBeacon signal sensing sequence, a user’s forward direction can be judged, in order to lead the user to the destination [7]. Yang et al. noted that an iBeacon set in medical locations can help users to find the proper diagnosis room or ward [26]. Ito et al. found that when a BLE beacon is arranged on a street, and a user’s mobile device scans the beacon signal, the information of merchants located in that area will be displayed [10]. Moreover, Faragher and Harle also proved the advantages of using BLE beacons for fingerprint positioning, and showed that significant positioning improvement over Wi-Fi is possible, even when using a small number of beacons [5]. These studies have all proved the feasibility of BLE beacon positioning, and show that positioning accuracy still has room for improvement.

In recent years, several well-known museums and researchers have developed mobile navigation systems with location-aware services; however, most of these systems only support outdoor GPS positioning function, and even when an indoor positioning function is provided, the range and accuracy are poor and location-aware services are limited. For example, the Discover NPM app, as launched by the Taiwan National Palace Museum, provides varied functions, such as traffic information, visitor guides, and route planning, and supports a GPS positioning function in the museum area, which allows users to know their position [4]. The American Museum of Natural History app from the USA can provide pictures and textual introductions about exhibits using the visitor’s location via Wi-Fi positioning and navigation of the paths inside the museum [3]. The app used for the Summer Palace in mainland China is also dominated by GPS, although position and navigation information can be obtained through iBeacon signal sensing. However, its positioning accuracy is insufficient, and navigation information is also limited [19]. The Taipei Fine Arts Museum App also applies the iBeacon Indoor Positioning System; when users arrive at the exhibition hall and receive the particular iBeacon signal, the App will automatically play the voice introduction of this exhibition hall; however, its positioning function is limited to area broadcasts, and cannot realize precise positioning. Compared with the system proposed by this paper, it has no AR recognition functions or display and navigation recommendations [18].

Table 1 presents the function comparison of a well-known museum’s mobile navigation system, which shows that many systems lack the indoor positioning function or precision positioning is limited. Moreover, most systems lack the recommendation function or only provide a simple navigation route recommendation function, but cannot provide personalized or effective navigation suggestions. In addition, most systems lack Markerless AR recognition and display functions, thus, they cannot provide more comprehensive 3D navigation information.

Table 1 Function comparison of a well-known museum’s mobile navigation system
Full size table

Overall, the above studies confirm the advantages and practicality of mobile devices combined with indoor positioning, recommendation, and AR display functions. However, the above-mentioned systems either lack an indoor positioning function or have a limited positioning function, and thus, cannot collect and analyze the user’s navigation behaviors for recommendation references. This paper proposes a mobile navigation system that integrates iBeacon indoor positioning and recommendation mechanisms, which could analyze the user’s navigation behavior and potential interests by detecting the user’s location and time, and then, actively recommend suitable exhibits and tour routes to the user for more systematic and efficient navigation. Moreover, accompanied by the markerless augmented reality display function, this system provides useful 3D and multimedia navigation information, in order to improve and enrich the user’s navigation experience.

This paper is organized, as follows. Section 2 introduces the main framework of the system. The proposed indoor positioning mechanism is discussed in Section 3. Section 4 describes the proposed recommendation mechanisms. The experimental results of indoor positioning are shown in Section 5. Section 6 presents the demonstration and discussion. Finally, concluding remarks are given in Section 7.

2 System overview

Figure 1 illustrates the system’s architecture, which is mainly composed of four parts: an AR navigation interface, an iBeacon indoor positioning system, a recommendation system, and a navigation information management system. The AR navigation interface is used to recognize and display 3D and multimedia navigation information; through markerless AR image recognition technology, the system can identify pictures displayed in the exhibition area using the mobile device, and then, display the corresponding 3D and multimedia navigation information on the AR navigation interface. In general, as basic navigation information is built into mobile devices, if the required navigation information has not been installed or downloaded into the mobile device, the system will proactively send an information demand, and then, download the required navigation information from the multimedia navigation information management system.

Figure 1
figure 1

Diagram of system architecture

Full size image

The iBeacon indoor positioning system is responsible for the real-time positioning of the user, meaning it can help the user identify their current position and find related exhibits faster and more easily. Moreover, the positioning system can identify specific users through the user identification function, thus, allowing the navigation server to provide specific data to specific users for personalized navigation. When the user actuates the mobile navigation system, the mobile device will first automatically link to the positioning server, then detect and send iBeacon signals to the positioning server. After completion of the positioning process, the mobile device will obtain the location information and navigation information from the positioning system and the navigation information management system, respectively.

The recommendation system provides information about exhibits that are consistent with the interests of the user. The recommendation system analyzes the questionnaire and tour behavior of the user, as well as the navigation records, by referring to the user database and exhibit database records, and then, recommends exhibits and tour routes that could interest the users. Meanwhile, it can record new recommendation routes to establish more reference data, which further improves the accuracy of the recommendation system.

The navigation information management system is mainly responsible for the transmission, reception, and integration of various navigation needs and information. The navigation server is responsible for receiving and integrating the requests of the recommendation system and obtaining relevant information from the multimedia database. Then, through the user identification function, specific data are sent to the mobile device at the user end to display the navigation information.

The flowchart of the system procedures is shown in Figure 2. When using this system for the first time, users are required to fill in their basic personal information and interests, which are then used as the basis for the analysis of the explicit behaviors by the recommendation system. Meanwhile, the system can preliminarily recommend exhibits and tour routes that are consistent with the interests of the user. In addition, the system can continuously record the location and time of the user through the positioning function, which is the basis for the analysis of implicit behaviors by the recommendation system, before updating the recommended exhibits and tour routes through integration with the original interests.

Figure 2
figure 2

Procedures and criteria of system implementation process

Full size image

After the navigation system receives the recommendation results, it can summarize and send back relevant navigation information to the user’s mobile device. The user can continuously assess the preferences of the exhibits during their visit, which allows more detailed analyses of the exhibits and interests, as well as real-time updating of follow-up exhibits and tour routes. Hence, with the increasing amount of assessment information, the accuracy of the recommendation system will be gradually improved.

3 Indoor positioning mechanism

Apple launched iBeacon in the summer of 2014, which features low power consumption, low cost, and good indoor positioning potential, thus, it can properly solve the problem of GPS being unable to realize accurate positioning indoors. Based on BLE, also known as Bluetooth 4.0 or Bluetooth Smart, iBeacon devices can create a signal area, and when a mobile device enters the area, the corresponding app will automatically sense the signal and execute the corresponding instruction.

Table 2 presents the comparison between the advantages and disadvantages of various indoor positioning methods. It can be seen that iBeacon has the advantages of low setup cost, high popularity, and feasible positioning accuracy and range. While Wi-Fi is popular, it still has poor accuracy in positioning. Conversely, while infrared ray and ultrasonic systems have good positioning accuracy, their positioning ranges are short, and they are easily influenced by light and signal attenuation. In addition, although Bluetooth (before 4.0) has the advantages of low setup cost and high popularity, its positioning range is too short, which limits its application scope. Moreover, while an active RFID device is a good choice, iBeacon is better due to its low setup cost, and the BLE protocol is supported by many mobile devices. Therefore, iBeacon has great potential in terms of popularity, and is likely to achieve wider application in indoor location-aware services.

Table 2 Comparison table of indoor positioning methods
Full size table

As the signal strength of iBeacon decreases with increased distance, on the basis of considering the environment and iBeacon’s individual differences, and according to the distance between a mobile device and iBeacon, this paper develops signal levels using different iBeacons for different distances. The method proposed in this paper could set an appropriate threshold value for individual label signals according to the individual differences of the environment and the hardware, thereby, effectively improving positioning accuracy and stability through signal grading and a signal level matrix (abbreviated as SLM) mechanism [26].

The iBeacon indoor positioning process of the proposed system is shown in Figure 3. First, prior to positioning, the positioning environment is laid out, and the reference values of the signal strength levels for the individual iBeacons at different distances are analyzed and set. Afterward, using these signal strength reference values, the SLM for the storage of all positioning areas is established in the database. During the real-time positioning process, the mobile device will input the collected signal level information to the SLM, and search for the same SLM in the database. In the case of an identical SLM, the corresponding areas will be directly displayed to complete positioning; if an identical SLM cannot be found, similar SLMs will be searched by the error tolerance mechanism to complete positioning; if an identical or a similar SLM cannot be found, the system will return to the first step of collecting signals for re-positioning.

Figure 3
figure 3

Schematic diagram of iBeacon indoor positioning process

Full size image

Figure 4 shows the SLM positioning principle given herein. One iBeacon with an interval of D meters is arranged in the positioning space, and each iBeacon (B11, B12, …, Bmn) is numbered, as based on the line coordinate. Using the location of the iBeacon as the center of a circle and D meters as the radius, the locating space is divided into numerous independent sensing areas. Thus, by analyzing the iBeacon’s signal level received by the mobile device, the positioning of the sensing area will be effectively completed.

Figure 4
figure 4

Diagram of SLM positioning principle

Full size image

For example, assuming the interval of D is 4 m, respectively, on the top, bottom, left, and right sides of the iBeacon, if the mobile device is very close to (less than one meter from) iBeacon B22, the strength of signal S22 sent by B22 will have signal level L0 (S22 = L0), while the signal levels of other iBeacons will be L3 (SLM is shown in Figure 5a). The location of iBeacon B22 can be directly positioned, and the location is named A22.

Figure 5
figure 5

Diagram of the SLM examples a SLM of A22; b SLM of A22U; c SLM of A22L; d SLM of A22UL; e SLM of A1122

Full size image

The areas of A22U, A22D, A22L, A22R, A22UL, A22UR, A22DL, and A22DR around A22 have signal levels of L2 or L3 when S22 = L1. For example, when S22 = L1, S12 = L2, and other signal levels are all L3, the A22U area will be positioned (SLM is as shown in Figure 5b). While, when S22 = L1, S21 = L2, and other signal levels are all L3, it will be positioned in A22L (SLM is as shown in Figure 5c). A22UL shows the situation when S22 = L1, S12 = L2, S21 = L2, and other signal levels are L3 (SLM is as shown in Figure 5d). In addition, when the SLM is as shown in Figure 5e, and the distance between the mobile device and all iBeacons exceeds 0.5D (2 m), the mobile device is in the middle of four iBeacons and can sense four L2 signal strengths, thus, it will be positioned in A1122 (the naming method is based on four iBeacons, A11, A12, A21, A22, and is the combination of the smallest and biggest numbers).

Algorithm 1 presents the SLM positioning algorithm of the system. First, the server will record the location of all iBeacons, then, based on the signal strength of the individual iBeacon and signal level correspondence table, the server will convert the received iBeacon signal strength into the signal level and save it into iBeacon_SLM_Map. Next, based on the signal level in the level matrix, the signal levels will be ranked from strong to weak and saved into SLM_Map. Finally, according to the intersectional location of the iBeacons that have strongest signal level, positioning can be rapidly and accurately completed.

figure a

4 Recommendation mechanisms

While indoor positioning technology is useful in museum mobile navigation, users still face the problem of an overwhelming amount of exhibit information, which makes it difficult to determine the exhibits and tour routes. Hence, it is important to add a recommendation system to the mobile navigation system, which could recommend suitable exhibits and tour routes after analysis, and according to the user’s behavior or active assessment information. Based on the concept of a recommendation system, this paper presents a mobile navigation system that could analyze the explicit behaviors of different users, including various items, such as personal interests and evaluations of the exhibits, as well as implicit behaviors, such as location and time, in order to recommend exhibits that may interest the user. Coupled with the positioning function, it could proactively recommend suitable exhibits and tour routes to facilitate navigation for the user.

This system supports four recommendation modes, including interest recommendation, hot recommendation, hybrid recommendation, and expert recommendation. Among them, interest recommendation refers to recommending the exhibits that may interest the user, which is determined according to the personal information and the interest options checked by the user. Hot recommendation refers to the ranking of the exhibits, as based on their visit frequency, to indicate popular exhibits and their visit rate; users can check the items they are interested in and a corresponding tour route will be automatically generated. The hybrid recommendation method is used to predict and recommend exhibits that may interest users, as based on their tour behavior. Expert recommendation refers to providing exhibits and tour routes planned by experts; users can choose suitable tour routes according to the estimated tour time provided by the system.

4.1 Interest recommendation

Interest recommendation is based on the content-based recommendation, which analyzes the correlation between the user’s interest and exhibits according to the user-selected personal data and interest options to further recommend exhibits that may interest the user. First, the system classifies exhibits according to the themes set by the Mackay Memorial Museum as an option of interest for the user to select, including: diary, document, photo, contract, cultural relic, etc. Next, the system analyzes the attributes and features of individual exhibits according to the themes displayed in each exhibition hall and showcased in Mackay Memorial Museum, and then, it analyzes the user’s personal interest and interest features according to the personal information provided the checked interest options; lastly, after the contrast of correlation between the user’s interest and exhibits’ attributes, the exhibits with higher correlation are recommended to the user.

4.2 Hot recommendation

The hot recommendation mode is based on the visit rate of all users of the exhibits. After systematic calculation and ranking, the list of popular exhibits is shown, which is used as the basis for the user to choose exhibits to visit. First, during the tour, the system will record the exhibits visited by users and add up the visit frequencies of the exhibits. Later, the number of visits of all users is analyzed to calculate the visit rate for all the exhibits, the exhibits with higher visit rates are list according to the ranking, and the visit rate is calculated after the exhibits become accessible. Meanwhile, the system enables the user to check items in the ranking of popular exhibits to further generate tour routes, and the system will continue to record, recalculate, and rank the visit rate of newly-added users.

4.3 Hybrid recommendation

In order to consider visitor’s explicit interests and potential interests, this system adopts hybrid recommendation, which is calculated by the sequential combination of content-based recommendations and collaborative recommendations, to predict exhibits that may interest the user. Hybrid recommendation has the advantage of content-based recommendations, including simplicity, effectiveness, and the ability to collect implicit behaviors, thus, it can overcome the issue of false behavior. Hybrid recommendation also has the advantage of collaborative recommendations to identify the potential interests of the user and avoid the problems of newly added data and the cold start issue, as based on the personal information obtained from content-based recommendations, allowing it to obtain diversified and accurate recommendation results [23].

The sequential combination method obtains the user’s interests and hobbies by content-based recommendations, and then, predicts new interests and habits through collaborative recommendation. First, a questionnaire survey is conducted to understand the user’s interests, which is used as a basis for the preliminary recommendation of exhibits and tour routes. Then, during the visit, the system continuously records the user’s preferences for various seen exhibits, and predicts the unseen exhibits that could possibly interest the users by evaluating the similarities between exhibits and the interest value of the user toward exhibits to recommend unseen exhibits and update the tour routes.

4.4 Expert recommendation

Some users may not have specific visit habits, and they do not want to refer to other people’s visit experience, but simply want to visit the most representative exhibits or the most classic tour routes. As the personnel of the museum have a good understanding of the exhibits in Mackay Memorial Museum, this system provides three classic tour routes with different lengths of time (10, 20, and 30 min) according to the suggestions of experts and the relevance of the exhibits, thus, visitors can rapidly choose tour routes designed by experts according to their visit time, and visit the most representative key exhibits and exhibition areas in an efficient and systematic manner.

5 Experiment results

The Mackay Memorial Museum, which is a Level 2 historical site in Taiwan, is used as the experimental site to test this system. Figure 6 shows the iBeacon layout in Oxford College for the testing of the indoor positioning function. First, based on individual exhibition room sizes and display cabinet positions, this study sets several iBeacons, respectively. The data transmitted by each iBeacon, as well as the signal strength of such data, are captured, and the UUID, Major, and Minor are taken as the identification basis of the individual iBeacons, in order to distinguish and analyze the data of the numerous iBeacon signals in the space.

Figure 6
figure 6

Layout of iBeacons in Oxford College

Full size image

Figure 7 presents the analysis diagram of the signal strength of different iBeacons measured at different distances, which shows that, although the iBeacon signal vibrates up and down, the signal strength is inversely proportional to the distance between the mobile device and the iBeacon. While the figure also shows that the signals at different distances are occasionally interlaced with each other, the proposed noise filtering mechanism could fix this problem appropriately.

Figure 7
figure 7

Diagram of the signal strengths of different iBeacons at different distances

Full size image

The proposed positioning system has two mechanisms to avoid the interference of signal noise and improve positioning accuracy. First, when the positioning system receives a new signal, it would compare this with the previous signal; if the difference of the two signals is more than two levels, the system would treat the new signal as invalid noise and directly filter it. Moreover, the positioning system adopts the moving average filter method to refine and smooth the signals.

Equation (1) shows the moving average filter method of real-time positioning, where x (t) represents the signal strength value received at time t, and x (t-1) represents the signal strength value received at time t-1... and so on, while y (t, n) denotes the average signal strength of the first n consecutive signals received at time t, which can be the reference value for real-time positioning. Figure 8 presents the analysis diagram of the filtered signal strength of different iBeacons at different distances, and the results indicate that the filtered signal strength is relatively stable and could effectively avoid the problem of signal strength interleaving. Figure 9 shows the average filtered signal strength of different iBeacons at different distances.

Figure 8
figure 8

Diagram of the filtered signal strengths of different iBeacons at different distances

Full size image
Figure 9
figure 9

Diagram of the average filtered signal strengths of iBeacons at different distances

Full size image
$$ y\left(t,n\right)=\left[x(t)+x\left(t-1\right)+x\left(t-2\right)+\dots +x\left(t-n+1\right)\right]/n $$
(1)

Based on the above test results and the signal level setup algorithm (as shown in Algorithm 2), and aiming at different iBeacons, this study sets up their respective suitable signal levels and corresponding signal strengths, as well as the SLM of all sensing areas (as shown in Table 3). Following this, the positioning system could collect the iBeacon signal strengths via a mobile device, and effectively calculate the user’s location based on the SLM. Moreover, all user-related location information is summarized, in order to analyze crowd distribution and flow in each exhibition area. Figure 10 shows the on-site test examples in Oxford College, as based on the SLM positioning mechanism.

Table 3 Table of signal levels and corresponding signal strengths for different iBeacons
Full size table
Figure 10
figure 10

Diagram of on-site test examples based on the SLM positioning mechanism

Full size image
figure b

Table 4 shows the analysis of the positioning accuracy rate of the experimental results. Based on the SLM positioning mechanism, in the case of using the original signal strength values of the iBeacons, accuracy is between 20% and 100%, with an average value of 75%. In the case of using the filtered signal strength values of the iBeacons, accuracy is between 50% and 100%, with an average value of 85%. In terms of accuracy under different positioning errors, which are practical for mobile navigation, the average accuracies of the positioning errors at one meter and two meters are 97% and 99%, respectively. The experimental results indicate that the accuracy of the proposed positioning mechanism is sufficient for the location-aware services of mobile navigation.

Table 4 Table of accuracy analysis
Full size table

6 Demonstration and discussion

During system implementation, Unity is used as the main development platform to implement the user interface of the mobile navigation system, and integrate the recommendation functions, indoor positioning, augmented reality recognition, presentation, network transmission, etc. The mobile device used for testing is ASUS Zenfone 3. Figure 11 shows the on-site test results in Oxford College and the navigation interface of the system. The navigation interface could be divided into three parts. The first part is the AR information display area in the middle of the interface; by aiming at specific pictures and texts in Oxford College, the user can detect and present related 3D and video navigation information. The second part is the text information display area in the lower middle of the interface, which displays the related text navigation information of the exhibition area, as well as the exhibits where the user is located. The third part is the position information display area in the lower right corner of the interface, which displays the user’s location on a small map of Oxford College, thus, allowing the user to easily know their location. During a visit, if users are particularly interested in certain exhibit items, they could use the mobile device’s camera to scan pictures of the exhibits and browse the augmented reality 3D and multimedia navigation information (as shown in Figure 12), thus, enriching and improving their navigation experience.

Figure 11
figure 11

Diagram of on-site test result and navigation interface

Full size image
Figure 12
figure 12

Diagram of demonstrating the detailed 3D navigation information

Full size image

Regarding the recommendation service, when using the system for the first time, users are required to fill in their basic information and select their interest options via the user interface, in order to produce the initial recommended tour routes (denoted by yellow lines, as shown in Figure 13). Users could also choose the hot recommended tour routes or expert recommended tour routes (as shown in Figure 14). Figure 15 shows a recommended example according to the hybrid recommendation. With the movements of the user, in addition to the initial yellow routes, the newly visited routes would automatically change to green to facilitate identification. Moreover, the system continuously records the visited locations and time as the basis for analysis of the implicit behaviors of the user, and updates the recommended new tour routes (red lines) individually.

Figure 13
figure 13

Diagram of the fill-in interface option of the interest recommendation and the corresponding tour route (yellow lines)

Full size image
Figure 14
figure 14

Diagram of the interface options of hot recommendation (left) and expert recommendation (right)

Full size image
Figure 15
figure 15

Diagram of updating the newly visited tour route (green lines) and the new recommended tour routes (red lines) of hybrid recommendation

Full size image

The proposed AR mobile navigation system supports indoor positioning and content recommendation services to enrich the user’s visit experience. In this study, 81 user questionnaires are adopted to investigate user satisfaction with the proposed mobile navigation system (including 52 university students who majored in computer science and 29 university students who majored in other areas). The questionnaire was designed using a Likert Scale, in which 5 points indicates strong agreement and one point indicates strong disagreement. In general, the average score was 3 ± 0.5 points, thus, this study defines an average score higher than 3.5 as agreement, while an average score less than 2.5 indicates disagreement. The results of the investigation are shown in Table 5; it can be seen that almost all of the average scores of the survey are higher than four points, indicating that most of the users are satisfied with the system functions, regardless of whether the users majored in computer science. While the score of Question 1, meaning users who did not major in computer science, is the lowest, it is still higher than 3.5 points (3.97 ± 0.72). This result indicates the need to develop a more user-friendly interface for general users in the future. Overall, this survey verifies the feasibility and practicability of the proposed mobile navigation system.

Table 5 Results of the questionnaire survey on satisfaction
Full size table

7 Conclusion

This paper presents an AR mobile navigation system that combines iBeacon indoor positioning and four kinds of content recommendation mechanisms. The proposed iBeacon positioning mechanism is able to flexibly set suitable sensing values for different iBeacon devices at different distances, in order to realize stable and accurate indoor positioning. Based on this iBeacon positioning mechanism, the developed mobile navigation system can accurately detect the visitor’s location, and promptly record the location and time of the visit, which can be used as a reference for content recommendation. The proposed recommendation mechanisms could predict and recommend exhibits and tour routes according to the interests and tour behavior of the user. In addition, accompanied by the markerless AR display function, this system could provide diversified 3D and multimedia navigation information. The results of the survey showed that users were satisfied with the system functions, and verified the feasibility and practicability of the proposed mobile navigation system.