Accelerating Emergency Response in Airport Environments: An Experimental Study on Intelligent Sound Detection Systems

This chapter introduces the critical importance of response times in emergency situations and highlights the role of intelligent systems in enhancing crisis management. By examining traditional methods and their limitations, the following sections provide a foundation for understanding the potential benefits of advanced technologies, such as Sound Event Detectors (SED), in improving response efficiency and public safety.

1.1 Context and importance of response times in critical incidents

The response time of security personnel and IRS units plays a crucial role in determining the outcomes of critical incidents in high-risk areas. Traditionally, security personnel and IRS units are alerted to these incidents and emergencies through notifications heavily influenced by human factors, which can result in significant delays. Individuals who could report incidents and emergencies are often the victims themselves. Human factors, such as delayed reaction or confusion in crowded environments, often hinder prompt reporting of emergencies. One known aspect is the 'bystander effect' [1], where individuals assume others have already reported the incident, but it is only one of many behavioural limitations in high-risk settings. Operational and administrative measures at airports highlight the importance of clearly defined procedures and minimizing delays in the transfer of information between IRS units [2].

1.2 Traditional reporting methods and their limitations

Traditional methods of reporting emergencies are often prone to errors caused by human factors. Intelligent sound event detectors (SED) offer an automated approach that recognizes specific learned sounds, such as gunshots, shouting, and glass breaking, thereby triggering faster responses from security forces and IRS units [3].

1.3 Benefits of intelligent sound event detectors

The importance of sound detectors in the field of security has grown significantly in recent years due to advancements in sound signal analysis technologies. For instance, prior research demonstrated the use of sound event detectors in transportation, achieving a significant improvement in response times to unexpected events [4]. These systems use machine learning models trained on large annotated audio datasets to classify predefined sound types (e.g., gunshots, screams) in real time. Their architecture typically includes front-end feature extraction (e.g., MFCC, Mel spectrograms), followed by classifiers like convolutional neural networks (CNNs) or support vector machines (SVMs) [5].

Another advantage of automated systems, such as SEDs, is their ability to eliminate human error. Research emphasizes that human factors often lead to errors in decision-making during crises, especially when individuals are exposed to stress or confusion [6]. Automation eliminates this element, which is particularly crucial during emergencies requiring a rapid response. Further studies document that smart sensors and data integration can significantly enhance the ability to respond quickly to crises, leading to better outcomes in emergency management [7].

Additionally, prior research has investigated the application of machine learning algorithms for sound classification, offering foundational insights into the effective analysis and categorization of audio signals [3]. Building upon these advancements, this study adapts such principles specifically to the context of emergency response and crisis management using Sound Event Detectors (SEDs). Recent investigations have also demonstrated the efficacy of support vector machines (SVM) in classifying indoor sounds, such as gunshots and breaking glass, which further enhances the accuracy and reliability of sound event detection (SED) systems in critical security applications [8].

1.4 Research objectives and hypotheses

The aim of this study is to evaluate the effectiveness of intelligent SED systems in reducing response times for security personnel and IRS units and their contribution to prevention and enhanced security levels. The research focuses on controlled simulations of an active shooter attack in the environment of an airport administrative building to provide a comparative analysis of traditional and automated reporting methods. We hypothesize that automated SED systems will serve as a significantly faster source of notification for security forces, thereby greatly improving response times and overall intervention efficiency.

The use of IoT technologies, such as SEDs, represents a critical advancement in crisis management, essential for reducing casualties and improving intervention efficiency [9]. While similar studies have explored automated systems in emergency contexts, to the best of our knowledge, no prior research has specifically focused on the integration of SEDs in the controlled simulation of active shooter scenarios within airport environments. This research aims to contribute to existing knowledge on the use of intelligent technologies for public safety and further develop methodologies that enable their effective integration into current security protocols. Systems like SED can play a key role in improving safety not only in administrative and commercial buildings but also in public spaces, where they can serve as the first line of defense during critical incidents [2]. Current research addresses key challenges including detection in noisy environments, distinguishing overlapping sounds, and reducing false alarm rates—especially in crowded public spaces. Emerging trends also include the integration of SED with video analytics and IoT frameworks to create context-aware, multimodal surveillance solutions. Despite these advancements, there remains a need for experimental validation of SED systems in complex, high-risk environments—especially within structured, time-critical scenarios such as active shooter incidents. This study aims to fill that gap.

The primary metric examined in this study is the time difference between notifications of an emergency event identified by sound event detectors (SEDs) and notifications received by security personnel at the operational center or the Police of the Czech Republic through conventional means. These conventional methods include a witness or participant calling by phone or using standard systems such as a traditional video surveillance system without sound detection, emergency buttons, etc. The results of the scenarios confirmed the hypothesis that sound detectors are generally faster.

3.1 Results of individual scenarios

The results of individual scenarios were analyzed using standardized procedures for evaluating the effectiveness of security systems [25]. The methodology for evaluating response times was based on recommended procedures for time-series analysis in security systems [26]. Statistical processing of data and interpretation of results followed a methodology for comparative analysis of security technologies [27], with a focus on the practical significance of the observed time differences for the efficiency of IRS interventions.

The tables below clearly demonstrate that automated systems provided a substantial acceleration in detection and notification, contributing to increased efficiency in IRS response.

Scenario 1: The emergency event began at 18:01:12 when a man approached the reception desk and verbally harassed the receptionist. This action was not accompanied by any loud sound. Notification to security and the monitoring center was carried out by pressing the emergency button on the receptionist's desk at 18:02:40. The time difference between the start of the event and the notification was 00:01:28. No notification via SED was made due to the absence of a sound event. A detection was only made much later upon the arrival of security personnel, when a verbal altercation occurred between the perpetrator and the security staff.

Scenario 2: The second scenario began at 18:43:45, when aggressive shouting was detected during a dispute between a man and the receptionist. The first notification via SED was sent at 18:43:50, a time difference of 00:00:05 from the start of the event. The first notification through traditional methods occurred when the receptionist called security at 18:44:15, representing a time difference of 00:00:30 from the start of the event. The difference in notification time between the two methods was 00:00:25. Later, at 18:46:45, security was notified again about the ongoing emergency situation.

Scenario 3: The third scenario began at 19:36:20 with the detection of gunfire at the first shot. The first notification via SED was sent at 19:36:25, a time difference of 00:00:05. The first notification through traditional methods occurred when an employee called the emergency line at 19:43:10, a time difference of 00:06:50 from the start of the event. The difference in notification time between the two methods was 00:06:45. Later detections were recorded on the second floor at 19:30:10 and 19:37:50.

Scenario 4: The fourth scenario began at 20:58:38, when gunfire was detected at the first shot. The first notification via SED was sent at 20:58:43, a time difference of 00:00:05. Traditional notification was made when an employee called the emergency line at 21:00:05, a time difference of 00:01:27 from the start of the event. The difference in notification time between the two methods was 00:01:22. A total of 10 shots and one instance of aggression were detected during the incident.

The difference between these two variants shows that automated detection via SED provided faster identification of the incident, leading to more effective IRS intervention. The time difference between the methods was 6 minutes and 17 seconds, with automatic detection providing the first alert of the incident at 20:58:43, while traditional reporting occurred several minutes later.

3.2 Analysed data displayed in tables

Table 1 summarises the timings for each scenario, which includes the start of the incident, the time of first notification by SED and the time of first notification by the standard route. The differences in notification times are also shown, allowing an evaluation of the speed and efficiency of the SED system compared to traditional methods.

Table 1 presents a comparison of emergency notification times, and Table 2 provides an overview of the sources of audio detection and the forms of first notification by the standard route in each scenario. It also contains notes describing the details of the progress of each incident. This table provides a better understanding of the circumstances of each scenario and how to respond to it.

Table 1. Comparison of emergency notification times

Table 2. Sources of sound detection and forms of emergency notification

Table 3. Statistical evaluation of the effectiveness of SED and standard notification

Table 3 provides a comparative overview of the effectiveness of SED and standard notification methods. It shows the mean, median response times and variability (standard deviation) of both methods. Each metric includes an explanation of its meaning and practical application. The table is designed to better understand the differences between the two methods and to evaluate the benefits of the SED system in emergencies

3.3 Additional findings

The first scenario was excluded from the evaluation as it did not include any prominent sound events, such as gunfire or aggressive shouting. It was excluded in accordance with standard data cleaning procedures for security research [28]. Results cleaned of this scenario better reflect the efficiency of the SED system compared to standard notification methods. However, the exclusion of the first scenario highlights one of the limitations of using SED—it is less effective in emergencies without prominent sound events. In practice, such events are less common or less critical, making SED most beneficial in detecting situations with clear sound manifestations.

3.4 Unexpected/negative findings

During the simulation, no false alarms were triggered by non-emergency sounds. However, this result might be influenced by the controlled nature of the experiment, as real-life environments may involve various natural sounds that could mimic gunshots or aggressive shouting. This highlights the importance of proper calibration to minimize false positives and ensure reliability in diverse conditions. The analysis of potential system limitations was conducted in accordance with methodologies for evaluating the reliability of security systems [29].

3.5 Summary of results

Based on the results from the three tables, the following conclusions can be drawn about the effectiveness of the SED system during emergencies:

(1) Significant Reduction in Response Time with SED: The average response time for the SED system was 5 seconds, whereas the standard notification time was 175.67 seconds. This represents a 97.2% reduction in response time compared to traditional methods. A two-sample t-test confirmed that the difference in response times between the SED and traditional notification methods was statistically significant (p < 0.01), indicating that the observed improvements were not due to random variation but reflect a robust effect.

(2) Consistent Performance of SED: The median response time and zero standard deviation indicate that the SED system provides consistent response times, meaning notifications were made without delays in all cases. In contrast, standard notifications exhibited significant variability (standard deviation of 204.93 seconds).

(3) Faster Detection and Notification: In individual scenarios (Table 2), the notification time using SED was always shorter than the standard notification time. The time difference ranged from 25 seconds to over 6 minutes, further confirming the speed advantage of the SED system.

(4) Qualitative Analysis of Scenarios: According to Table 3, SED effectively detected sound events such as gunfire and aggressive shouting, leading to faster responses. Traditional notification methods, such as calling an emergency line or pressing a button, were slower and showed greater variability in response times.

Overall, the results demonstrate that the SED system is an effective tool for rapid and consistent notification of emergencies, particularly those accompanied by sound manifestations. This confirms its benefit in enhancing security measures and reducing the response time of security forces.

The integration of intelligent sound event detectors (SEDs) significantly enhances the efficiency of IRS units by reducing response times. The findings of this study support the implementation of advanced sound detection technologies in high-risk environments, such as airports, to improve security outcomes. Automated detection enables rapid alerts to IRS units within seconds of an incident, which is crucial for minimizing damage and protecting lives. These conclusions align with recent research highlighting that IoT-based systems and advanced algorithms can achieve detection accuracies of up to 85.4% and significantly improve crisis management efficiency.

This study demonstrated that intelligent SED systems can identify emergencies faster than traditional methods reliant on human factors. The difference in response times, ranging from several seconds to several minutes, profoundly impacts the efficiency of IRS interventions. In specific scenarios, such as active shooter incidents, automated detection shortened the time required to identify and apprehend attackers, leading to more effective interventions and reduced casualties. Similar multi-sensor system approaches have been successfully applied in urban areas, underscoring their practical utility in emergency management.

Although no false alarms appeared during the experiment, it is important to bear in mind the controlled nature of the study and the potential for varied background noises in real-world environments. These factors highlight the need for further calibration and refinement of detection algorithms to ensure system accuracy and reliability. Regular calibration and maintenance of SED systems remain crucial for their proper functionality, especially when applied in complex, noisy conditions. Regular testing across diverse environments is recommended to sustain the system’s high reliability.

Recent studies have demonstrated that using noise-agnostic multitask learning, particularly integrating acoustic noise profiling into detection models, can reduce false alarm rates without impairing accuracy. Furthermore, new architectures based on transformers—previously dominant in language processing—are now proving highly effective in audio detection. These models offer improved generalization in acoustically complex environments and reduce dependency on environment-specific calibration, making SED deployment more scalable and adaptable.

Based on the findings, we recommend deploying SED in critical locations such as airports, shopping centers, and public buildings, where rapid response can significantly enhance safety. Combining SED with other technologies, such as camera systems and motion detectors, ensures comprehensive protection even in scenarios lacking sound cues. This combination increases the coverage and efficiency of emergency detection.

Personnel operating these technologies should receive regular training on using SED and interpreting their outputs. This includes security staff and IRS units to ensure seamless responses to notifications from SED.

Future research should explore other types of intelligent sensors and their combined impact on emergency response efficiency. The integration of additional technologies, such as security cameras, thermal sensors, or motion detectors, could further enhance detection capabilities, thereby improving safety and security measures. Additionally, the development of more advanced machine learning algorithms could contribute to greater accuracy and reliability of detection systems, minimizing false alarms and improving overall effectiveness.

Pilot deployment of SED in various environments, such as industrial facilities, schools, and healthcare facilities, could help identify the specific needs of each environment and ensure optimal system configurations. Such deployments could also contribute to a better understanding and effective use of these technologies in practice.

One of the novel contributions of this study is demonstrating how SED systems can fill a critical gap in existing crisis management architectures. While most surveillance relies heavily on visual input, many high-risk incidents begin with audio cues—such as gunshots or aggressive shouting—that may occur out of camera view or without distinct visual indicators. By adding 'ears' to the 'eyes' of traditional surveillance systems, SED enables faster, more context-aware incident detection. Multimodal systems that integrate audio and video data have shown superior accuracy and relevance in identifying threats compared to unimodal approaches [41]. This integration enhances the effectiveness of the entire security chain—from deterrence and detection to information transfer and coordinated response.

The overall findings of this study clearly demonstrate that intelligent sound detection systems have the potential to significantly improve the response time of security forces, such as facility security and IRS units, thereby enhancing public safety in high-risk environments. This technology is particularly valuable for prevention and timely response. The implementation of these technologies could represent a significant step forward in crisis management and population protection.