AES New York 2019
Paper Session P11
P11 - Semantic Audio
Thursday, October 17, 1:15 pm — 2:45 pm
Robert C. Maher, Montana State University - Bozeman, MT, USA
P11-1 Impact of Statistical Parameters of Late Reverberation on the Instantaneous Frequencies of Reverberant Audio—Sarah R. Smith, University of Rochester - Rochester, NY, USA; Mark F. Bocko, University of Rochester - Rochester, NY, USA
This paper addresses the impact of late reverberation on the instantaneous frequency tracks of reverberant audio. While existing models of early reflections and low frequency room modes enable prediction of instantaneous frequency tracks of a filtered signal, the effects of late reverberation are best modeled statistically. After reviewing the parameterization of late reverberation, the effects of frequency dependent decay time and direct to reverberant ratio on instantaneous frequency are investigated using synthetic impulse responses derived from velvet noise. These effects are quantified using the autocorrelation function of the reverberant instantaneous frequency tracks. Finally, the instantaneous frequency deviations that occur when an anechoic sound is filtered with a recorded impulse response are compared to those resulting from synthesized late reverberation.
Convention Paper 10279
P11-2 Precise Temporal Localization of Sudden Onsets in Audio Signals Using the Wavelet Approach—Yuxuan Wan, Hong Kong University of Science and Technology - Clean Water Bay, Hong Kong; Yijia Chen, Hong Kong University of Science and Technology - Clean Water Bay, Hong Kong; Keegan Yi Hang Sim, Hong Kong University of Science and Technology - Clean Water Bay, Hong Kong; Lijia Wu, Hong Kong University of Science and Technology - Hong Kong, Chian; Xianzheng Geng, Hong Kong University of Science and Technology - Clean Water Bay, Hong Kong; Kevin Chau, Hong Kong University of Science and Technology - Clean Water Bay, Hong Kong
Presently reported is a wavelet-based method for the temporal localization of sudden onsets in audio signals with sub-millisecond precision. The method only requires O(n) operations, which is highly efficient. The entire audio signal can be processed as a whole without the need to be broken down into individual windowed overlapping blocks. It can also be processed in a streaming mode compatible with real-time processing. In comparison with time-domain and frequency-domain methods, the wavelet-based method proposed here offers several distinct advantages in sudden onset detection, temporal localization accuracy, and computational cost, which may therefore find broad applications in audio signal processing and music information retrieval.
Convention Paper 10280
P11-3 Forensic Comparison of Simultaneous Recordings of Gunshots at a Crime Scene—Robert C. Maher, Montana State University - Bozeman, MT, USA; Ethan Hoerr, Montana State University - Bozeman, MT, USA
Audio forensic evidence is of increasing importance in law enforcement investigations because of the growing use in the United States of personal audio/video recorders carried by officers on duty, by bystanders, and by surveillance systems of businesses and residences. These recording systems capture speech, background environmental sounds, and in some cases, gunshots and other firearm sounds. When there are multiple audio recording devices near the scene of a gunfire incident, the similarities and differences of the various recordings can either help or hamper the audio forensic examiner’s efforts to describe the sequence of events. This paper considers several examples and provides recommendations for audio forensic examiners in the interpretation of this gunshot acoustic evidence.
Convention Paper 10281