The reliability of New Haven’s ShotSpotter system, a series of sensors which detect and locate gunshots in the city’s central neighborhoods, has come under question after several missed gunshots and false reports in recent weeks.

After the system of audio detectors, which uses acoustic triangulation to register gunshots, failed to report gunfire in a number of recent cases — including the city’s first homicide of the year on Saturday in Dixwell — the New Haven Police Department is reviewing its continued use. Though NHPD spokesman David Hartman said the department has not decided whether to retain or tweak ShotSpotter, he added that some of the inaccuracies are attributable to the “human-like” errors of the system because it must hear gunshots in the same way as witnesses.

“ShotSpotter is quite human, in that not only does it hear things, but it has to make an analysis of what it hears,” Hartman said. “When something goes pop, bang or boom, the system has to register it and them determine what kind of pop, bang or boom it was, and sometimes it doesn’t get it right.”

ShotSpotter was launched in the Elm City September 2009 and comprises over 20 sensors that were installed with funds from a $500,000 grant from the Justice Department.

When a sensor detects a loud noise and registers it as a gunshot, the three sensors nearest it locate the shot to within four feet and alert NHPD dispatchers.

Recently, however, ShotSpotter has failed to pick up several gunshots and even registered several instances of nonexistent gunfire. Hartman said the system recorded no gunshots in or around the Dixwell neighborhood, where the Elm City’s first homicide victim of the year was found Saturday morning at the intersection of County and Munson streets.

“Then again, there were also no earwitness reports either — it could either be a fault of the system or the gunfire could have taken place in a vehicle, where it wouldn’t be picked up,” he said.

Physical structures and other environments that cause echoes can hamper the system’s ability to detect gunshots, the company’s official website notes.

In response to the department’s reliability concerns, representatives from ShotSpotter’s Mountain View, Calif., headquarters came to the NHPD’s Union Avenue headquarters last month to explain a newly developed protocol to prevent false alarms.

Data from ShotSpotter will now be directed to a company-wide diagnostic system, which can differentiate between gunshots and other possible acoustic triggers, before the information is passed to the NHPD’s 911 response center. Before this change, reports from the ShotSpotter sensors were sent directly to dispatchers.

A 2011 report by the company titled “ShotSpotter Gunshot Location System Efficacy Study” acknowledges that “false positives” — ShotSpotter reports that do not come from gunshots — are the “single most common complaint” of users and pose an operational problem.

The report found that 67 percent of alerts on average are caused by actual gunfire.

Still, NHPD Chief Dean Esserman told the New Haven Register he wanted to talk to ShotSpotter officials again to ensure that all questions about the system’s performance are resolved. He said his department needed to “either gain confidence in ShotSpotter” by resolving the reliability issues or shut it down.

Hartman noted that despite the accuracy issues, the system has proven useful in reporting many shootings that go unreported by earwitnesses.

“If we can’t get this system to work to its full potential, then one train of thought is, ‘Why support this system?'” he said. “The other train of thought is that you’re going to get some false readings in anything like this, and if you consider the one false reading with 11 positive readings, do you throw away the dozen eggs when one’s cracked?”

ShotSpotter systems operate in more than 45 cities around the United States, including Los Angeles and Boston.