AI School Anti-Bullying Early Warning Solution
Back to Solutions

AI School Anti-Bullying Early Warning Solution

Dihu AI School Anti-Bullying System — deep integration of acoustics and language models, AI keyword recognition of bullying behavior, 24/7 intelligent protection of campus safety, accurate recognition of local dialects. According to statistics, it has covered 27 provinces, with sales of 150,000+ sets, 2,500+ partner schools, and 135 covered cities.

Highlights

Built-in 58 fixed keywords + 15 custom keywords, with built-in alarm keywords such as Help, A fight, Someone is hitting someone, Stop hitting, etc.
20-second pre-recording before alarm, fully preserving the cause and entire dialogue process of the bullying incident, forming complete traceable audio evidence
24-hour uninterrupted voiceprint detection + automatic keyword alarm, with false alarm rate controllable within 10%
POE/DC 12V dual power supply, wired / 4G/WIFI multiple networking methods, adaptable to wiring conditions in different scenarios, low construction difficulty
Long-distance intercom distance ≥2m, 5W high-quality speaker for remote shouting to deter, can be connected to an external pickup, sound and light alarm light, and ≥20W high-power horn
The device mainboard is coated with conformal coating; the device appearance is rainproof, moisture-proof, and corrosion-proof, with a long service life, suitable for complex campus deployment environments
Remote batch configuration and upgrade, greatly reducing operation and maintenance workload and labor costs
Customizable high-incidence alarm periods (class breaks, lunch breaks, after school), including alarm logs and multi-condition filtering, making it convenient for schools to analyze high-incidence areas and periods.
Specially optimized for multiple dialects such as Jiangsu-Zhejiang, Southwest, Northwest, and ethnic minority languages, with a built-in massive bullying keyword library and support for importing local custom lexicons.
Recognition of ≥300 generic speech phrases, accurately identifying bullying behaviors such as verbal abuse and conflicts, physical assault, and calls for help, while intelligently distinguishing the voice characteristics of perpetrators, bystanders, and victims, with precise noise reduction and ultra-low false alarms.
Intelligent noise/loudness detection alarm; customizable decibel threshold and sound duration; when the on-site sound pressure level meets the set conditions, the device automatically triggers an alarm.
Equipped with a localized AI computing chip, built-in multi-microphone array pickup and Dihu self-developed noise reduction algorithm, and adopts dual intelligent judgment logic based on duration + decibels, which can accurately filter out invalid environmental sounds such as water flow, fan noise, and noisy crowds during class breaks.
Multi-terminal collaborative alarm reception, supporting synchronous alarm reception on the server side, management host, web client, agent platform, mobile APP, WeChat Mini Program, and DingTalk client, following a first-answer-first-handling mechanism; when one terminal answers, the other terminals are automatically locked to avoid duplicate handling.
Self-developed platform with core software and hardware, supporting both SIP and proprietary protocols, with strong openness and compatibility; it can be connected via SDK to mainstream education security platforms on the market, supports ONVIF video linkage, smoke sensors, radar, infrared, and other third-party sensing hardware, enabling joint analysis and judgment by multiple devices at multiple points and expanding the dimensions of security warning.
Supports remote batch configuration of device parameters, remote upgrades, audio self-test, and power-off recovery; batch configuration of 1,000 devices at one time greatly reduces construction and commissioning labor; comes with the Android Xiaodi commissioning assistant APP for scan-code configuration, repair reporting, viewing device parameters, and remote commissioning.

Advantages

1
Products are available in visual and pure-audio non-visual versions, achieving full coverage of all campus scenarios, strictly complying with laws and regulations on the privacy protection of minors, eliminating privacy leaks and public opinion risks, and making student privacy control autonomous and controllable.
2
Visual terminals can be selected for public areas such as dormitory corridors and public hallways, providing dual assurance of audio warning + video verification; pure-audio terminals without video capture should be selected for privacy-sensitive areas such as restrooms and dormitory cubicles.
3
One-click switch between silent mode and false-alarm prevention mode. Supports silent alarms (no sound or light prompts at the front end) and one-click activation of false-alarm prevention mode, balancing false alarms and missed alarms to avoid sound and light alarms disrupting normal teaching order
4
It also provides dedicated noise control for dormitories after lights-out, suitable for nighttime discipline management in boarding dormitories. The platform name, alarm rules, and keyword library can all be customized
5
One set of equipment simultaneously supports anti-bullying alerts, emergency calls, two-way intercom, zone broadcasting, and dormitory noise control, enabling reuse across multiple scenarios with high cost-effectiveness
6
Local LAN deployment: wired and PoE devices are networked together with a local management host and server, suitable for schools with isolated intranets and data that must not leave the campus
7
Graded alarm push notifications + synchronized pop-ups and one-click voice broadcast deterrence
8
Edge AI local analysis, regular audio is not uploaded, supports private deployment
9
7×24-hour scheduled broadcasting, zone broadcasting, TTS text-to-speech broadcasting; management personnel can make remote announcements and intervene in on-site conflicts in a timely manner.
10
Adopts a brain neural network AI speech recognition processor, with a dual-engine architecture of offline NLP recognition + cloud-based large model in-depth analysis
11
Internet cloud deployment: 4G/wired terminals, cloud servers, no local server required, fast implementation and launch
12
Supports MicroSD card local independent storage, caches recordings when disconnected, and automatically uploads after reconnecting to ensure data is not lost
13
Multi-microphone array + proprietary noise reduction + dual judgment based on duration and decibel level, low false alarm rate
14
Full-process closed-loop handling + bullying data statistics and alarm heat map analysis
15
Native SIP protocol + seamless integration with same-source systems, reusing existing low-voltage systems
16
AI + silent SOS dual-mode alarms, with specialized optimization for multiple dialects
17
Flexible deployment modes, suitable for different project scales, with two deployment architectures

Detailed Content

Solution Positioning

Targeting primary and secondary schools, secondary vocational schools, and boarding campuses, and addressing the four core pain points of campus bullying: "high concealment, difficulty in actively seeking help, difficulty in evidence collection and responsibility determination, and delayed response," it builds a full-scenario on-campus prevention and control system of active early warning - rapid intervention - closed-loop handling - data analysis and assessment, while strictly adhering to the compliance bottom line of minor privacy protection and data security.

It conducts advance prevention of potential bullying incidents and initiates voice call alarms, provides timely intervention and stopping of ongoing bullying incidents, and performs post-event analysis and verification of bullying incidents that have already occurred, so as to better prevent and stop campus bullying incidents and effectively improve the efficiency and level of campus safety work.

Solution Architecture

Solution Architecture

Large Model Analysis and Determination

  • Can reduce the false alarm rate of device alarms and increase accuracy to 90%

Large Model Analysis and Determination

Alarm Process

Alarm Process

Get Solution

Contact our professional team for detailed solution materials and quotation