LiveTranStack

Description

• Hardware / Software solution to enable live interpretation of talks, anywhere there's IP and power (we could use power banks as USB UPSs) Could be simplified as a simple IP intercom / duplex HW-SW stack

Goal

• At least two mic/headphones
• Cheap
• OSS / OSHW
• Qualitatively good

Obstacles

• Network, software mixing and codec latency
• Mics getting ambient noise

Components

• Raspberry pi or other GNU/Linux embedded system with USB, ethernet and GPIO
• pots, switches and LEDs
• USB mic-headphones
• OGG Opus
• Power source
• [2 line lcd?]

HW details

We need a mute momentary switch. Also a buzzer (to diagnose start-up / majors problems)

All of what follows will be implemented twice par device (or plus). USB port for the audio devices, and a way to be sure which is which. Two (digital?) pots (mic and headphone gains) A enable/disable switch (with built-in led if cheap enough?) One LED to indicate if a mic is open (if not inside switch).

We map the digipots to alsa volume controls.

SW details

Then, let's assume a GNU/Linux distro

Setup: Parameters such as audio source (upstream) IP:port, also downstream IP:port and codec parameters are stored somewhere (sd) Also the language code of the stream (can we mux this somehow, or is it upstream's responsiblity)

Optionally drive a 2line led display and have enough buttons to set ip:ports and number of devices?

On boot: Enumerate USB devices. If one is absent, keep beeping its led. If both are absent, buzz till reset. If an upstream is set, attempt connection. If no connection, buzz / blink in another pattern. Software mix both sources and feed the output into an OGG Opus encoder. Stream this (UDP broadcast? or set ip:port) (Optionally mix upstream with lower level)

While true: while mute button is pressed: Cut USB inputs. Don't change upstream level. If no input is enabled, Upstream level = 1; If at least one input is enabled, Upstream < 1; On upstream failure: attempt reco (maybe double beep and blink) On downstream failure: same (maybe triple beep and blink)

Roadmap

• Test streaming a mono Opus to a raspi, look at usage, figure out a couple quality / latency settings.
• Test encoding a mono Opus from a raspi, same.
• Test doing the same with software mixing enabled (broadcasting a mix of the input stream and some local audio)
• Test doing the same using an USB audio device as playback and capture at the same time.
• Double it.
• Figure out how to GPIO efficiently at the same time (implement buzzer and visual alerts)
• Figure out how to control audio volumes via GPIO
• Test the darn thing in a real setting
• Monitor power use, derive estimation of battery-operation capacity

Participants

• User:Fzn (Leader)