ESP32-P4/C5 Multimodal WebRTC Doorbell

00 / Context & objective

Embedded audio and video, radar and visual presence sensing connected to browser calls and a servo door-lock demonstration.

01My contributions

• Adapted OV5647 video, ES8311 two-way audio and the ESP32-C5 wireless link on the WT99P4C5-S1 board.
• Integrated LD2450 radar and visual presence detection with zone checks, event debouncing and repeated-ring suppression.
• Connected a PWM servo queue, hold and reset sequence to the Chinese browser interface and call interactions.
• Adjusted LAN connection waits and media cleanup, and organized device logs and acrylic manufacturing assets.

02System flow

1. 01Radar and visual sensing
2. 02Visit and ring policy
3. 03Browser WebRTC session
4. 04Video and two-way audio
5. 05Servo unlock demonstration and reset

03Recorded results

• Historical logs record three LAN call-setup times of about 4, 6 and 4 seconds, compared with about 30 seconds initially.
• Existing demonstrations cover browser calls, video, audio and servo motion; structural manufacturing assets are also public.
• Archive records show 14 doorbell check entry points and 17 structural-design tests passing, all on the host side.

04Validation scope & limitations

Read results alongside their measurement targets and conditions. Historical records, host checks, and device validation are distinguished.

• WebRTC and media foundations come from Espressif. My work covers board adaptation, sensing extensions, interaction and integration.
• A trusted-LAN, single-browser prototype without user-level unlock authentication. An unloaded servo simulates the lock without real lock feedback.
• The 4, 6 and 4 second observations are three historical trials, not evidence of Internet performance or long-term reliability.
• Archiving did not rebuild or flash firmware or repeat device acceptance. The 8 kHz call and 16 kHz prompt-mixing paths retain a compatibility issue.

05Public materials