Page 27 - Special Issue
P. 27

 Forensic Acoustics
 moment to moment. By maintaining a database of the small, instantaneous aleatoric variations of the grid frequency for every date and time, the precise hum frequency present in an evidentiary recording can be matched to the database. The degree to which the re- corded hum matches the database pattern can either confirm or refute the claimed date and time that the recording was made. Nonetheless, even in a system in- volving ENF detection and comparison, an examiner may not be able to rule out the possibility that a clever adversary filtered out the presence of any residual ENF. What’s more, a skilled audio forger could conceivably synthesize a fake ENF signal and additively mix it into the evidentiary signal, saving the contrived composite information back onto the digital storage medium as a way to fool a subsequent forensic examiner into believ- ing that the recording is authentic.
Figure 3. Example of a time waveform (top) and spectrogram (bottom) de- picting an urban surveillance recording containing gunshots, reverberation, mechanical sounds, yelling, car horns, and other background sounds and noise. Several aspects of the recorded soundscape may be of interest in a par- ticular investigation, such as the sequence and timing of audible events and any intelligible speech utterances.
Many digital audio recording and storage systems in- corporate metadata in the digital file format. Metadata
may include information about the recording settings, date and time, manufacturer of the device, and its software ver- sion. Although metadata can potentially be altered to con- ceal tampering with the audio data contained in the file, an audio forensic examiner should always review the metadata as part of an authenticity investigation (Koenig and Lacey, 2014).
Other evidence of digital audio tampering can be more sub- tle, requiring careful consideration of signal continuity and the interpretation of background sounds. An attempt to edit out an embarrassing or possibly incriminating speech utter- ance, for example, could cause a momentary, but detectable, interruption of the distinctive background sounds present in the original recording (Maher, 2010).
Innovation in Acoustical Forensics
Many key principles of audio forensics have been developed and evolved incrementally over the last 50 years, but cer- tain aspects of the field gain increasing interest from time to time. Here are a few of the current “hot topics” in foren- sic acoustics that could entice ASA members to join the re- search effort.
Audio Scene “Fingerprinting”
As noted previously, a forensic recording typically includes prominent foreground sounds that are of interest to the in- vestigation, such as speech utterances, warning alarms, or perhaps gunshots (National Academy of Sciences, 1982; Maher, 2007; Beck et al., 2011). Yet it is the background 26 | Acoustics Today | Summer 2015
sounds and environmental acoustics at the recording scene that can sometimes become an even more important part of the forensic evidence. The reverberation and background acoustic emissions can provide valuable information about the authenticity of the recording or help determine the geo- metric orientation of the sound sources with respect to the microphone. Typical forensic recordings involve a compli- cated superposition of many different sounds and the effects of microphone and audio encoder peculiarities (Figure 3).
In audio forensics, the term acoustical fingerprinting re- fers to the analysis of background sounds, acoustic reflec- tions, reverberation, and peculiarities of the microphone and recoding system that are detectable in the evidentiary recording (Alexander et al., 2012; Moore et al., 2014). Be- cause many common consumer recording devices include automatic gain control and perceptual audio coding/data compression algorithms intended for speech signals, addi- tional research is needed to understand the degree to which the acoustical surroundings and the characteristics of the recording microphone and digital audio-coding algorithms can be derived from the stored audio recording.
Proliferation of Personal Audio/Video Recording Devices
Technology for portable audio/video recorders has pro- gressed to the point that audio forensic evidence from a crime scene may come from one or more mobile smart- phones or cameras carried by journalists or bystanders, se- curity surveillance systems, dashboard-mounted recorders in police cars, and, increasingly, pocket-sized personal video
Spring 2020, Special Issue | Acoustics Today | 27 Reprinted from volume 11, issue 3

   25   26   27   28   29