Abstract
This paper presents a counseling (ro)bot called Visual Counseling Agent (VICA) which focuses on remote mental healthcare. It is an agent system leveraging artificial intelligence (AI) to aid mentally distressed persons through speech conversation. The system terminals are connected to servers by the Internet exploiting Cloud-nativeness, so that anyone who has any type of terminal can use it from anywhere. Despite a promising voice communication interface, VICA shows limitations in conversation continuity on conventional 4G networks. Concretely, the use of the current 4G networks produces word dropping, delayed response, and the occasional connection failure. The objective of this paper is to mitigate these issues by leveraging a 5G/6G slice inclusive of mobile/multiple edge computing (MEC). First, we propose and partly implement the enhanced and advanced version of VICA. Servers of enhanced versions collaborate to increase speech recognition reliability. Although it significantly increases generated data volume, the advanced version enables a recognition of the facial expressions to greatly enhance counseling quality. Then, we propose a quality assurance mechanism using multiple levels of catalog, as well as 5G/6G slice inclusive of MEC, and conduct experiments to uncover issues related to the 4G. Results indicate that the number of speech recognition errors in Internet Cloud is more than twofold compared to edge computing, implying that quality assurance using 5G/6G in conjunction with VICA Counseling (ro)bot has higher efficiency.
Introduction
Remote mental healthcare is a growing necessity, especially during emergencies like pandemics when human counselors are overwhelmed. AI-driven bots like the Visual Counseling Agent (VICA) offer a scalable solution. However, therapeutic "sensitive listening" requires immediate, accurate responses to build rapport—a "flow" that is often broken by the limitations of current network infrastructure. Conventional 4G networks suffer from "word dropping," high latency (often over 10 seconds), and frequent connection failures. This paper explores how next-generation 5G/6G architectures, specifically network slicing and Mobile Edge Computing (MEC), can move AI processing closer to the user to make AI counseling viable.
System Architecture and Methodology
The researchers developed VICA through several iterations, starting from CRECA (a text-based baseline) to VICA Basic (voice-enabled) and finally Advanced VICA, which incorporates multi-sensor fusion for facial expressions and heart rate monitoring.
As shown in the system architecture, the processor handles input voice through an input script that interacts with a speech recognizer (like Cloud Speech API). The core counseling agent then performs morphological analysis and response generation. To ensure quality, a novel catalogue-based mechanism was proposed with four tiers of service quality. Depending on the situational urgency, the system automatically requests specific network "slices" (e.g., dedicated bandwidth or MEC) to maintain performance.
The dialog processing logic involves an analyzer and a context generator that feeds into a context-based reasoner. This reasoner utilizes counseling knowledge, including regular matching rules and procedural knowledge, to generate appropriate responses. Furthermore, error recovery algorithms using perplexity checking and n-gram methods are employed to validate speech-to-text results.
Experimental Evaluation
The system was evaluated using four Japanese subjects (ages 30-70) and a set of 11 difficult-to-recognize Japanese sentences involving complex grammar or emotional distress. Performance was compared between a standard Internet Cloud setup and an Edge Computing setup using 4G networks in Tokyo. Key metrics included response times and the categorization of speech recognition errors (disappeared words, disappeared phrases, and wrong words).
Results and Discussion
The experiments revealed that Edge Computing (MEC) significantly improves both speed and accuracy. Speech recognition errors in the Internet Cloud were more than twofold compared to edge computing (e.g., 25 dropped phrases in the Cloud vs. 12 at the Edge). By processing data closer to the user, the system avoids the packet loss and jitter typical of the open internet.
Under 4G congestion, response times frequently exceeded 10 seconds, which is considered "fatal" for maintaining therapeutic continuity. Qualitative surveys confirmed this, with 80% of users reporting significant irritation with 4G-related delays. Counseling is a unique use case for 5G/6G because it is more sensitive to latency than many other high-bandwidth applications.
Conclusion
While the voice-based VICA is functional, the full implementation of the "Advanced" version with facial expression fusion remains future work. The study concludes that proactive network slicing and MEC are essential for providing the ultra-low latency and high reliability required for effective AI-driven mental healthcare. Future directions include exploring 6G's potential for "medical grade" sensor fusion and deeper emotional analysis.
Summary updated