A Review of Artificial Intelligence (AI) Algorithms for Sound Classification: Implications for Human-Robot Interaction (HRI)
Abstract
This report presents a review of artificial intelligence (AI) algorithms and their application to audition in a human-robot interaction (HRI) context. The AI algorithms selected for auditory perception ultimately have an impact on computational transparency, system behavior explain ability, and ultimately, the quality of the HRI. AI algorithms applied to auditory perception include sounds sensed and processed by a software system, as well as sounds emitted by a software system that are meant to be recognized by a human listener. Some major classes of AI algorithms, specifically neural networks, deep learning, hidden Markov models, and hybrid models will be reviewed in the context of machines sound processing. Additionally, the effects of each class of algorithm on transparency and HRI will be discussed. Recent work in AI algorithm development suggests that hybrid models may be the best approach for sound processing as they are recommended for complex data processing and decision-making. Hybrid models blend approaches to maximize the benefits while minimizing the limitations of multiple techniques. A set of general recommendations are included in the final section of the report.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 23, 2020
- Accession Number
- AD1090606
Entities
People
- Kelly Dickerson
- Troy Kelley
Organizations
- United States Army Research Laboratory