Gunshot Detection Technology
How networks of acoustic sensors detect and locate suspected gunfire, how algorithms and human review distinguish gunshots from other impulsive sounds, and what agencies should understand about alert reliability, dispatch, officer response, evidence, forensic reports, community impact, workload, privacy, performance measurement, procurement, and governance.
What this explainer does
Acoustic gunshot detection systems use networks of microphones or acoustic sensors to identify loud impulsive sounds that may represent firearm discharges. When multiple sensors capture a qualifying sound, software can analyze the acoustic event and estimate where it originated.
Depending on the system, machine classification may be followed by human review before an alert is transmitted to law enforcement. The resulting alert can give dispatchers and officers a location, time, number of suspected rounds, and other incident information before anyone calls 911.
The potential benefit is straightforward: police may learn about gunfire that otherwise would not be reported and can respond more quickly to injured victims, offenders, witnesses, firearms, cartridge cases, vehicles, video, and other evidence.
The limitation is equally important: the system is detecting and classifying sound. An alert does not by itself prove who fired a weapon, whether a firearm was actually discharged, whether the reported location is exact, or whether criminal conduct occurred.
Gunshot detection technology (GDT) is the general technology category.
ShotSpotter is a commercial product currently offered by SoundThinking and is one of the best-known implementations of outdoor acoustic gunshot detection.
Agencies should distinguish the technology category from a particular vendor, product, configuration, contract, or performance claim.
1. Overview
Gunshot detection technology converts outdoor sound into a machine-assisted public-safety alert.
Traditional police awareness of gunfire depends heavily on human reporting: a victim, witness, officer, or nearby resident hears or observes gunfire and contacts 911.
Acoustic gunshot detection adds a second reporting mechanism. Sensors positioned across a defined coverage area listen for acoustic events that meet system thresholds. When multiple sensors detect an event, software can estimate its origin and classify whether the sound is consistent with gunfire.
That capability can help police respond to gunfire that was never reported by a member of the public.
2. How Acoustic Gunshot Detection Works
3. What Does a Gunshot Sound Like to a Sensor?
A firearm discharge generates a short-duration, high-energy acoustic event commonly called a muzzle blast. Depending on firearm, ammunition, environment, and projectile characteristics, additional acoustic features may also exist.
To a computer system, however, the relevant question is not simply whether a sound was loud.
Classification can consider acoustic characteristics such as amplitude, duration, waveform, frequency information, impulsiveness, the relationship among multiple detected sounds, and other system-specific features.
Explosive sounds can resemble gunfire acoustically.
Sudden combustion or mechanical noises may produce similar impulsive sound.
Nail guns, machinery, metal impacts, and other industrial sounds can create challenging acoustic events.
Weather can generate powerful impulsive or complex acoustic signals.
Buildings and other surfaces can reflect sound and complicate interpretation.
Traffic, aircraft, crowds, music, wind, and dense urban noise can interfere with detection.
4. How the System Estimates Location
Sound travels through air at a measurable speed. If several geographically separated sensors detect the same acoustic event, the sound reaches those sensors at slightly different times.
By comparing those arrival-time differences and known sensor locations, the system can estimate where the sound originated. This process is commonly described as triangulation or, more precisely in some configurations, multilateration.
5. Machine Classification and Human Review
Modern gunshot-detection systems may use algorithms or machine-learning models to determine whether an acoustic event is sufficiently consistent with gunfire to warrant further processing.
Some systems also use human acoustic reviewers who evaluate candidate incidents before an alert is sent to police.
Software determines whether sensor data meet thresholds for a potentially relevant impulsive event.
Algorithms evaluate whether the acoustic characteristics resemble gunfire rather than another sound.
Where provided, trained reviewers may evaluate candidate incidents before police receive the alert.
6. What an Alert May Contain
Estimated coordinates, address, or mapped area associated with the acoustic event.
Timestamp showing when the acoustic event was detected.
Estimated number of separate impulsive events classified as suspected shots.
Short acoustic recordings or waveform information associated with the detected event may be available.
System-specific classification or review information may describe the incident as probable gunfire.
Alerts can be integrated with dispatch, mapping, RTCC, video, ALPR, or other public-safety systems.
7. Officer Response to a Gunshot Alert
An acoustic alert may justify sending officers to investigate, particularly because injured victims or physical evidence may be present even when no 911 caller reports the event.
But the alert does not erase ordinary constitutional rules governing detention, search, seizure, entry into private property, or use of force.
8. What an Alert Does—and Does Not—Prove
| Possible Inference | What the Alert Actually Establishes |
|---|---|
| “A gun was definitely fired.” | The system classified an acoustic event as suspected or probable gunfire according to its process. |
| “The shooter stood exactly here.” | The system estimated the acoustic origin within its reported location parameters. |
| “This person fired the gun.” | The alert does not identify a shooter. |
| “A crime occurred.” | Firearm discharge may be criminal, lawful, accidental, or misclassified; further investigation is necessary. |
| “There were exactly five shots.” | The system detected and classified a number of impulsive acoustic events; physical evidence may produce a different count. |
| “No alert means no shooting.” | Detection systems may miss gunfire because of coverage, acoustic, technical, environmental, or classification limitations. |
9. What Does “Accuracy” Mean?
Accuracy claims can be misleading if the metric is not defined.
Gunshot detection has several distinct performance questions.
Of actual firearm discharges within coverage, how many does the system detect?
Of detected acoustic events, how accurately does the system distinguish gunfire from other noises?
How close is the estimated origin to the actual source?
How much time passes between the acoustic event and delivery of a usable law-enforcement alert?
How frequently does responding police activity find independent evidence consistent with a shooting?
How frequently do alerts lead to victims, evidence, investigative leads, arrests, or other defined public-safety outcomes?
10. False Positives and Unconfirmed Alerts
A false positive occurs when the system classifies a non-gunshot sound as gunfire.
An unconfirmed police response is different. Officers may fail to locate evidence even though gunfire actually occurred.
Cartridge cases can be removed. A shooter may use a revolver. Evidence may fall in inaccessible locations. Victims may leave. Witnesses may not cooperate. Gunfire may occur outside the officer's search area. The reported acoustic coordinate may be imperfect.
Conversely, some alerts unquestionably can result from non-gunfire acoustic events.
11. False Negatives and Missed Gunfire
The opposite error is equally important: a firearm is discharged within or near a coverage area but no actionable alert is produced.
The shooting occurs outside effective sensor coverage.
Buildings, dense construction, or acoustic shadowing can affect sensor reception.
A loud environment can interfere with acoustic detection.
Firearm characteristics and suppression can affect acoustic signature.
Software or human review may reject an event that was gunfire.
Equipment failure, positioning, calibration, connectivity, or maintenance can affect performance.
12. What Research Has Found
Research on gunshot detection does not support a simple conclusion that the technology is either universally effective or universally ineffective.
NIJ-sponsored evaluations have found that implementation can increase police awareness of firearm incidents, affect response, and improve opportunities for evidence recovery in some settings.
Other findings have shown limited or inconsistent effects on shootings, crime reduction, case clearance, or other ultimate outcomes.
Sensors can identify suspected gunfire that was not otherwise reported to police.
Faster or more precise response can increase opportunities to recover cartridge cases or other scene evidence.
Detecting gunfire is not itself a violence-prevention intervention, and research has not consistently shown reductions in shootings.
13. Officer Workload and Opportunity Cost
A technology that detects previously unknown incidents necessarily creates additional police work.
That may be a benefit: previously undiscovered victims or evidence can receive attention.
But it also means agencies must evaluate the opportunity cost of sending officers to repeated alerts that do not result in confirmed criminal incidents.
How many officer-hours are consumed responding to alerts?
How long do officers spend looking for victims, witnesses, casings, property damage, or other evidence?
What calls, patrol functions, investigations, or community activities are delayed because resources are committed elsewhere?
14. Gunshot Alerts as Investigative Evidence
Gunshot-detection information can potentially assist investigators with chronology, estimated location, round count, acoustic sequencing, and identification of areas where physical evidence may exist.
It can also be correlated with:
Physical evidence can confirm that firearm discharge occurred.
Surveillance, BWC, doorbell, traffic, or RTCC cameras may show the event or associated people and vehicles.
Vehicle detections near the relevant time and place may generate investigative leads.
Witness reports can independently corroborate time, sound, direction, or observed activity.
Firearms examination and cartridge-case comparison may connect separate shooting scenes.
Injuries and treatment records may establish an actual shooting and timing.
15. Forensic Reports and Courtroom Use
Some gunshot-detection providers can produce incident reports, acoustic records, maps, timing information, and expert testimony for criminal proceedings.
When those materials become part of the prosecution, agencies and prosecutors should understand the complete analytical chain.
| Record | Why It May Matter |
|---|---|
| Raw Sensor Audio / Data | Underlying acoustic information associated with the event |
| Machine Classification | How software initially characterized the sound |
| Human Review | Whether a reviewer confirmed, rejected, or modified classification |
| Location Calculation | How the estimated acoustic origin was derived |
| Alert Record | What information law enforcement actually received |
| Revision History | Whether location, round count, classification, or other information changed |
| System Version | Software or algorithm configuration used at the time |
| Maintenance Records | Potentially relevant to sensor or network performance |
16. Legal Issues During Police Response
The legal issue with gunshot detection is often not whether police may receive an acoustic alert. The more immediate issue is what officers do because of it.
Presence near an alert location does not automatically make every person reasonably suspected of criminal activity.
Proximity alone does not automatically establish individualized reasonable suspicion concerning every nearby vehicle.
An acoustic coordinate near a residence does not by itself erase the warrant requirement for entry into the home.
Officers must apply governing reasonable-suspicion standards rather than automatically frisking everyone in the alert area.
A gunshot alert is context, but force must remain justified by the circumstances confronting officers.
Additional evidence is ordinarily necessary to connect a specific person to firearm possession or discharge.
17. Acoustic Privacy
Outdoor gunshot-detection systems necessarily use microphones or acoustic sensors, which can create public concern about whether police or vendors are continuously recording conversations.
Agencies should understand and be able to explain the system's actual audio architecture rather than relying on generalized assurances.
18. Community Impact and Deployment Geography
Gunshot-detection sensors are generally deployed selectively rather than uniformly throughout an entire jurisdiction.
That means placement decisions determine which communities receive both the benefits and burdens of increased automated police awareness.
Detection may direct first responders to injured people when nobody called 911.
More alerts can produce more high-risk police responses and encounters in covered neighborhoods.
Concentrating sensors in selected neighborhoods can create significantly different levels of technology-assisted police observation across a city.
19. Procurement and Contract Metrics
Gunshot detection is commonly purchased as a service covering a defined geographic area. Contract design can strongly influence how performance is measured.
| Contract Issue | Agency Question |
|---|---|
| Coverage | Exactly what geographic area is covered and how are gaps identified? |
| Detection Performance | How is a true gunshot established for purposes of measuring detection? |
| Location Accuracy | What distance threshold defines contract compliance? |
| Alert Speed | From what event is response latency measured? |
| False Alerts | How does the contract define and measure false positives? |
| Missed Gunfire | How can false negatives be identified and audited? |
| Sensor Uptime | What availability and maintenance requirements apply? |
| Data Access | Does the agency receive raw data, incident records, audit logs, and export capability? |
| Model Changes | Can algorithms or review procedures change without agency notice? |
| Independent Audit | May the agency or an independent evaluator test contract performance? |
| Evidence Support | What forensic reports, expert testimony, and litigation support are included? |
| Termination | What happens to historical acoustic data and agency records when the contract ends? |
20. How Agencies Should Evaluate Performance
Agencies should measure the technology against their own operational objectives instead of relying exclusively on vendor-wide performance statistics.
21. Agency Governance Framework
Define how alerts may be used operationally and what independent facts are required for enforcement action.
Specify alert priority, information relayed, call coding, cancellation, and coordination with 911 reports.
Train officers on what the technology detects, its limitations, tactical response, and constitutional boundaries.
Preserve alert information separately from officers' independent observations at the scene.
Develop consistent dispositions for confirmed gunfire, evidence found, no evidence, canceled alerts, and other outcomes.
Investigate significant false alerts, missed shootings, or location discrepancies.
Establish procedures for preserving acoustic records and forensic information when used in an investigation.
Address authentication, expert testimony, discovery, algorithmic records, and vendor evidence before contested cases.
Define audio retention, access, permitted use, and prohibitions on unrelated monitoring.
Independently verify performance metrics rather than relying only on vendor reports.
Explain coverage areas, objectives, capabilities, privacy rules, costs, and evaluation criteria.
Reassess whether documented operational benefits continue to justify cost, workload, deployment, and community impact.
22. Questions Every Agency Should Answer
23. Where Gunshot Detection Is Going
Machine-learning models may improve differentiation among gunfire, fireworks, vehicles, construction, and other impulsive sounds.
Alerts can automatically cue nearby cameras, mapping, dispatch systems, and real-time intelligence platforms.
Systems may automatically identify or reposition cameras near the estimated source location.
Vehicle detections near an acoustic event may be rapidly surfaced to investigators.
AI may combine acoustic events with CAD, 911, video, ALPR, BWC, and investigative databases.
Acoustic systems may increasingly be incorporated into broader smart-city, campus, transportation, or critical-infrastructure networks.
24. Key Terms
25. Related ShieldPST.ai Resources
Networked vehicle detection, historical searches, retention, alerts, and investigative use.
Open resource →AI-assisted video review, transcription, search, redaction, generated reports, and evidentiary integrity.
Open explainer →Preservation, metadata, discovery, authentication, and evidentiary integrity.
Open resource →Using AI to evaluate and correlate investigative evidence while preserving human verification.
Open resource →Research Fourth Amendment and emerging police-technology decisions.
Browse case library →Return to the Shield Technology Reference Library.
Browse explainers →26. Selected Primary and Authoritative Sources
NIJ-sponsored evaluation and implementation guidance based on research examining gunshot-detection technology across multiple jurisdictions.
Review NIJ publication
BJA guide addressing basic acoustic gunshot-detection principles, implementation, research, community engagement, and operational best practices.
Review BJA guide
NIJ-supported evaluation examining implementation and effects of acoustic gunshot detection across multiple police departments.
Review research brief
Multi-city research examining effects of gunshot-detection deployment on police awareness, workload, and response.
Review NIJ research
Research examining whether gunshot detection affected evidence collection and investigative outcomes.
Review research
Multi-method evaluation examining crime, response, evidence, and operational outcomes associated with gunshot detection.
Review study
Independent municipal analysis examining CPD responses, criminal-case reports, investigatory stops, and operational effects associated with ShotSpotter alerts.
Read OIG report
Independent audit examining performance monitoring, confirmed shootings, alert outcomes, contract administration, and sensor coverage.
Review audit
Current agency policy describing NYPD's use of acoustic gunshot-detection technology, system purpose, data, safeguards, access, retention, and operational use.
Review NYPD policy
Current manufacturer information concerning ShotSpotter's acoustic sensors, detection workflow, human review, alerting, and operational capabilities.
Review manufacturer information
27. Key Takeaways
- Gunshot detection uses networks of acoustic sensors to identify, classify, and locate sounds that may represent firearm discharges.
- Location is generally estimated by comparing when the acoustic event reaches multiple sensors.
- Modern systems may combine automated classification with human acoustic review before an alert reaches law enforcement.
- A gunshot alert is an investigative lead—not conclusive proof that a firearm was discharged.
- An alert does not identify the shooter, prove a crime, establish an exact firing position, or independently establish reasonable suspicion or probable cause concerning every person in the vicinity.
- “Accuracy” must be defined. Detection rate, classification accuracy, location accuracy, alert latency, police confirmation, and operational value are different metrics.
- A police response where no evidence is found should not automatically be labeled a technical false positive.
- Conversely, agencies must account for actual false positives and for firearm discharges that fail to generate alerts.
- Research indicates that gunshot detection can increase police awareness of firearm incidents and may improve response or evidence opportunities, but effects on shootings and ultimate investigative outcomes have been mixed.
- Increased detection also increases officer workload, creating an opportunity cost that should be measured.
- Agencies should distinguish sensor information from officers' independent observations in reports, affidavits, and testimony.
- When acoustic data or forensic reports become evidence, agencies should preserve relevant source information, classification history, review activity, location calculations, and system records.
- Policies should make clear that constitutional standards for stops, searches, home entries, seizures, and force continue to apply after an acoustic alert.
- Agencies should understand what audio sensors capture, what is retained, who can listen to it, and whether unrelated use is technically possible or prohibited.
- Deployment geography should be transparent and based on defensible criteria because sensor placement determines where automated gunfire detection and resulting police responses occur.
- Contract performance should be independently evaluated against agency-defined objectives rather than relying only on vendor performance claims.
- The next major governance challenge is sensor fusion: gunshot alert → RTCC → camera → ALPR → intelligence search → officer response. Agencies should evaluate that combined workflow as a system.