A Novel Approach to Listener Performance Modeling for Aircraft Detection in Urban Environments

ABSTRACT: In this proposal, we first extend the notion of speech reception threshold (SRT) to aircraft sound reception threshold (ASRT). In speech recognition, SRT refers to the mixture signal to noise ratio (SNR) required to achieve a certain intelligibility score, typically 50%. SRT is used to quantify speech intelligibility under various SNR conditions. Here we define ASRT as the SNR required to achieve 90% correct detection of aircraft sounds. ASRT will be used to quantify the capability of aircraft sound detection under various SNR conditions. Second, we propose to model humans"capability on aircraft sound detection using ideal binary mask (IBM). An IBM is defined as a binary matrix within which 1 denotes that the target (aircraft sound) energy in the corresponding time-frequency (T-F) unit exceeds the interference (non-aircraft sound) energy by a predefined threshold and 0 denotes otherwise. It was found that the IBM plays an important role in human intelligibility. By the same token, the IBM will likely to play a substantial role in aircraft sound detection. Third, it is also important to quantify the listening model by connecting the IBM with AFRT under different noisy environments. Extensive experiments will be performed in Phase 1. BENEFIT: The proposed listener model has great potential in providing insights about human"s listening capability of aircraft sounds under various noisy conditions. We will produce a software tool for our model. It can also be used for other military and civilian applications. For example, detection of submarine using sonar, detection of acoustic events for perimeter defense, etc. are some other notable applications. The combined market for the above mentioned applications in this paragraph will easily exceed 10 million dollars. This is based on license cost of $2000 and a market size of 5,000 systems.