While audio analytics can be very capable, there are some cases where performance may not be optimal. “In very noisy places like busy nightlife scenes, you will always have false detections,” Van der Vorst said. “As with other new technologies, it is important to set the expectations of clients right.”
Some clients think that every alarm would be a real fight or at least people shouting aggressively. This is not the case, as there are some other sounds that can trigger alarms as well, Van der Vorst cautioned. “While stressed human voices and nonhuman sounds can be reliably distinguished, it is very challenging for current algorithms to tell the difference between a group of drunk people laughing loudly (stressed voices) in a busy nightlife area and someone or a group of people shouting aggressively.” Both voices will trigger an alarm, but car horns, construction noises, ambulance sirens and many other loud nonhuman sounds will not be mistaken for signs of aggression.
A computer will never be as smart as the human brain, so there will always be alarms that are irrelevant to the operator, Van der Vorst said. “Audio analytics is an early warning system, but you always need human verification to see if you really need to take action. As long as the audio analytics system is integrated well into the operators' work flow, it will only take a few seconds to verify an alarm.”
In addition, prospective buyers must understand that there is no magic involved in the creation of this technology. There are clients that try to cover a large area with a small number of sensors; when using only a few sensors, only a limited area can be covered, Van der Vorst said. The flip side of the coin is clients that want to try the system in only a very limited area, automatically missing out on incidents that are outside the covered area.
Some of the newer values and opportunities created through the application of audio analytics are very exciting, but will take a while for integrators to become fully aware and well-versed in what can be achieved, Mitchell said. “Wider adoption of audio analytics is likely to be triggered through firsthand experiences of system integrators and end users gaining familiarity and confidence in what the technology can deliver.” Continued customer feedback and competition are always welcome, as they help steer future developments of new solutions that can detect and protect a wider range of environments. Increased functionality and reliability will allow audio analytics to be deployed in a greater range of applications, thereby snowballing market growth.