Audio-Based-Attendance System

An innovative attendance system that uses voice recognition to mark attendance efficiently and securely. This project leverages audio processing and speech recognition to streamline the attendance process in classrooms, workplaces, or events.

---

Features

• Speech Recognition: Converts spoken input into text to mark attendance.
• Voice Authentication: Validates identity using voice biometrics (optional).
• Real-Time Processing: Instantly logs attendance.
• Secure Data Storage: Stores attendance data in a database with encryption.
• Accessible: Suitable for visually impaired individuals or hands-free environments.

---

Technology Stack

• Programming Language: Python
• Audio Processing: SpeechRecognition library, pyannote.audio
• Database: Supabase (PostgreSQL backend)
• Frameworks: Flask or Streamlit (for UI)
• Cloud Services (optional): Google Speech-to-Text, Azure Speech Services
• Hardware: Microphone or audio input device

---

Workflow Diagram

Below is a GitHub-friendly textual representation of the workflow:

1. Professor's Laptop: Audio is recorded in real-time.
2. Audio Submission: The professor submits the audio to a secure database (encrypted).
3. Students' Laptops: Students retrieve the professor's audio after submission.
4. Audio Comparison: Comparison is performed locally on students' laptops between the live room audio and the submitted professor's audio.
   • Match Found: Attendance is marked.
   • No Match Found: Attendance is denied.
5. Dashboard Update: Attendance records are updated for both students and professors.

Animated Workflow (Steps Visualization)

+-------------------+
| Professor Laptop  |
| (Audio Recording) |
+-------------------+
        |
        v
+-------------------+
| Submit Audio to   |
| Database (Encrypt)|
+-------------------+
        |
        v
+-------------------+
| Students Retrieve |
| Audio for Local   |
| Comparison        |
+-------------------+
        |
        v
+-------------------+
| Compare Live Room |
| Audio with        |
| Submitted Audio   |
+-------------------+
     /       \
    v         v
+-------+   +--------+
| Match |   | No     |
| Found |   | Match  |
| Mark  |   | Deny   |
| Attend|   | Attend |
+-------+   +--------+
        |
        v
+-------------------+
| Dashboard Update |
+-------------------+

---

How It Works

1. Professor Audio Submission: Professors record live audio on their laptops and submit it to the database, where it is stored in encrypted form.
2. Student Audio Comparison: Students retrieve the professor's submitted audio and compare it with live audio recorded on their local devices.
3. Attendance Marking: Attendance is marked based on the audio comparison result.
4. Proxy Mitigation:
   • VPN proxies are detected and mitigated.
   • Sharing pre-recorded audio is ineffective as live audio differences are detectable.
   • The system ensures privacy without using ML/DL models or biometric data.

---

Application

Login page

1. The login process for student is different here. It checks if the user is logging in from the regular device or not.
2. if not, then student should go through the forgot/generate process where the request will be sent to admin to approve the login from new device and then only new device will be registered and old device ID will be deleted.
3. This process ensures that user can do secure login from one device.

Professor Dashboard

Student Dashboard

---

Mitigating

1. VPN proxy will mitigated.
2. sharing the recorded audio will be mitigated as the application can find the difference in live audio and recorded audio.
3. It is much more efficient than OTP-based attendance.
4. NOT using ML/DL models just student biometric information privacy.

---

Future Enhancements

• Noise-cancellation for improved accuracy.
• Analytics dashboard for attendance trends.

---