PJM/IA – Police Interview Emotion Analysis System
Overview
PJM/IA is an intelligent support platform designed to assist criminal investigators during interviews and interrogations through multimodal emotion analysis.
The system combines multiple Artificial Intelligence approaches to analyze emotional behavior from:
- Facial expressions (FER – Facial Emotion Recognition)
- Voice tone (SER – Speech Emotion Recognition)
- Natural language sentiment analysis
In addition to emotion analysis, the platform automates the administrative workflow associated with interviews by generating official reports automatically and producing annotated multimedia evidence.
Main Features
Emotion Recognition
- Facial Emotion Recognition (FER)
- Speech Emotion Recognition (SER)
- Sentiment Analysis using NLP
Automatic Documentation
- Automatic generation of:
- Complaint Reports
- Witness Examination Reports
- Defendant Interrogation Reports
- Automatic PDF report generation
- Automatic transcription of interview audio into text
Multimedia Processing
- Video analysis
- Audio analysis
- Live camera streaming analysis
Evidence Generation
- Annotated interview videos
- Emotion timeline tracking every 10 seconds
- Structured interview archive system
System Architecture
The application is divided into modular components:
| Module | Description |
|---|---|
| FER | Facial expression emotion recognition |
| SER | Voice emotion analysis |
| NLP | Sentiment analysis from speech transcription |
| Document Engine | Automatic report generation |
| Streaming Engine | Real-time camera processing |
Supported Inputs
Video Formats
.mp4.avi.mov
Audio Formats
.mp3.wav.ogg
Technical Requirements
| Component | Minimum | Recommended |
|---|---|---|
| RAM | 8GB | 16GB |
| Operating System | Windows XP | Windows 10+ |
| Camera | USB Camera | USB Camera |
| Microphone | USB Microphone | USB Microphone |
| Storage | 5GB | SSD Recommended |
Installation
Windows Installation
Run:
PJM_IA_Installer.exe
Follow the installation wizard until completion.
The installer will:
- Install the application
- Create the required directories
- Generate a desktop shortcut
Authentication
Default credentials for testing purposes:
Username: kl3z
Password: 12345
Usage
1. Load Subject Information
The investigator must first load:
- Citizen Card (CC) or
- Military Registration Sheet (FM)
This information is mandatory.
2. Configure Interview Metadata
Optional metadata:
- Interviewer
- Date
- Time
- Location
- City
3. Choose Analysis Mode
Video Analysis
Processes prerecorded interrogation videos.
Audio Analysis
Processes audio recordings.
Live Streaming
Processes live camera input in real-time.
Press:
Q
to stop recording.
Generated Output
After processing, the system generates:
- Official PJM Word reports
- PDF summary report
- Full interview transcription
- Emotion analysis reports
- Annotated video files
All files are stored automatically in:
/Corpus/<CC>_<DATE>_<TIME>/
Example:
12345678_2026-05-20_15-30-00
Interview Analysis
The system performs emotion tracking every 10 seconds and correlates:
- Facial expressions
- Vocal tone
- Speech sentiment
This allows investigators to identify emotional variations throughout the interview.
Technologies Used
- Python
- OpenCV
- NLP
- Speech Recognition
- Machine Learning
- Facial Recognition
- Emotion Classification
- Audio Signal Processing
Research Context
This project was developed as an AI-assisted criminal investigation support tool focused on:
- Interview enhancement
- Behavioral analysis
- Automated documentation
- Multimodal emotion recognition
Disclaimer
This project is intended for research and investigative support purposes only.
Emotion recognition systems should not be considered definitive indicators of deception or guilt.
Human supervision and professional investigative procedures remain essential.
Future Improvements
- GPU acceleration
- Real-time dashboard
- Advanced speaker diarization
- Multi-camera support
- AI-assisted interrogation insights
- Deepfake detection integration
License
Private / Research Project