🎓 Farewell, Berkeley! 🌟
As of December 15, 2023, my Berkeley journey has come to an end. It's been an incredible two-year ride, and I'm truly honored to have had the opportunity to run this account. But, as they say, all good things must come to an end. 🌆
With my UC Berkeley graduation now behind me, I'm bidding farewell to the Bay Area and heading back to the sunny streets of Los Angeles. 🌞 Thanks for all the memories, Berkeley – you'll always have a special place in my heart! ❤️
If you wish to stay up to date, you can follow my LA account on Twitter or connect with me on LinkedIn.
📚 Explore My Berkeley Collection 📸📜📽️
Crime/Dispatch Codes/Common Terms
Photographs/Videos
Public Record Requests
📰 Raw OIS Footage 📰
Details are in the description of this YouTube video.
Details are in the description of this YouTube video.
Details are in the description of this YouTube video.
Details are in the description of this YouTube video.
🔧 System Technical Insights 🔍
The @ScanBerkeley radio and transcription system relied on four open-source software tools and a combination of hardware to deliver its functionality.
- Trunk-Recorder: This tool was used to record each individual transmission sent over the Berkeley talk groups. It essentially captured all radio communications.
- Rdio-Scanner: To play each transmission as it came in and allow listeners to customize their experience, I used Rdio-Scanner. This gave users the ability to toggle and set the scanner according to their preferences.
- Trunk-Transcribe: After recording the transmissions, Trunk-Transcribe came into play. It processed these recorded files by running them through OpenAI's Whisper API, converting the audio content into text.
- Modified Version of crimeisdown-v3: Finally, a highly customized version of crimeisdown-v3 was utilized to showcase the transcriptions on a webpage and facilitate historical review.
- SDRs (Software-Defined Radios): These are essential for receiving radio signals. We used 3x Nooelec RTL-SDR v5 Bundle, which you can find here.
- Raspberry Pi 4: The Raspberry Pi served as the central processing unit for Trunk-Recorder and Rdio-Scanner. It handled the recording and playback of radio transmissions. We used the Raspberry Pi 4, available here.
Unlike traditional scanners that save audio as a single file, in this system, each transmission was saved as its own file. This approach offered the flexibility for users to select what they wanted to listen to by simply clicking a button.
Hardware Components
It's worth noting that the Raspberry Pi 4 is a relatively modest machine for this task, and it's generally recommended to use a more robust system, such as an old desktop computer, if you plan to handle a larger volume of voice recorders (transmissions) simultaneously. While I did experience a system crash on one occasion due to excessive voice activity, in normal circumstances, it managed to handle all Berkeley channels without any issues.
To ensure seamless operation, the entire setup was configured to run behind an NGINX proxy. This proxy served as the hosting platform for both the API and website systems, making them accessible within my local network.
Hosted Frequencies
| DEC | HEX | Mode | Alpha Tag | Description | Tag | Group | Priority |
|---|---|---|---|---|---|---|---|
| 2105 | 839 | D | Berkeley FD1 | Fire Dispatch 1 | Fire Dispatch | Berkeley | 1 |
| 2106 | 83a | D | Berkeley FD2 | Fire Dispatch 2 | Fire Dispatch | Berkeley | 1 |
| 2112 | 840 | D | Berkeley FD Emer | Fire Emergency | Emergency Ops | Berkeley | 1 |
| 2671 | a6f | D | ACFD Command 11 | Command 11 - Berkeley | Fire-Tac | Alameda County Fire | 1 |
| 2672 | a70 | D | ACFD Command 12 | Command 12 - Berkeley | Fire-Tac | Alameda County Fire | 1 |
| 2691 | a83 | D | ACFD Tac 31 | Tac 31 - Berkeley | Fire-Tac | Alameda County Fire | 1 |
| 2692 | a84 | D | ACFD Tac 32 | Tac 32 - Berkeley | Fire-Tac | Alameda County Fire | 1 |
| 2711 | a97 | D | ACFD Tac 51 | Tac 51 - Berkeley | Fire-Tac | Alameda County Fire | 1 |
| 2712 | a98 | D | ACFD Tac 52 | Tac 52 - Berkeley | Fire-Tac | Alameda County Fire | 1 |
| 3105 | c21 | T | Berkeley PD1 | Police Dispatch | Law Dispatch | Berkeley | 1 |
| 3106 | c22 | T | Berkeley PD2 | Police Ch 2 | Law Tac | Berkeley | 1 |
| 3108 | c24 | T | Berkeley PD4 | Police Ch 4 | Law Tac | Berkeley | 1 |
| 3110 | c26 | T | Berkeley PD6 | Police Investigations 1 | Law Tac | Berkeley | 1 |
| 3605 | e15 | T | UCB PD1 DSP | Police Dispatch | Law Dispatch | UC Berkeley | 1 |
| 3606 | e16 | D | UCB PD2 | Police Ch 2 | Law Tac | UC Berkeley | 1 |
| 3608 | e18 | D | UCB PD4 ENF | Police Enforcement | Law Tac | UC Berkeley | 1 |
| 3609 | e19 | D | UCB PD5 | POLICE TAC | Law Tac | UC Berkeley | 1 |
However, please note that channels BPD3 (also known as RB1), BPD5, BPD7, and UCPD6 were not available as they were encrypted.
Transcribe System Overview
The transcription and historical site were hosted on a DigitalOcean droplet, utilizing Docker for containerization. The monthly cost for this setup was approximately $35.
Audio storage was managed through Amazon Web Services (AWS) S3, where approximately 200,000 transmissions were stored over a few months. The cost for S3 storage varied but typically amounted to around $2 per month.
Additionally, the transcription system employed OpenAI Whisper, incurring a cost of approximately $50 or more per month. It's worth noting that Whisper can also be run locally for free if you have access to an extra graphics card and other hardware.
One valuable feature of the Transcribe system was its ability to send notifications to my phone and Slack in real-time. These notifications were triggered the moment a phrase matched my configuration file. This functionality allowed me to tune in instantly when a priority incident was detected.
Software Configuration Files