Can a noise-adaptive voice gate fit in about 120 bytes of RAM on a pendant microcontroller, and does a cheaper noise floor cost battery life?

What it found

A stylised follow-up to ambient-gateroc. Its adaptive energy gate (open when a 20 ms frame is at least 4× the minimum energy of the last 5 s) is ported to plain C99 and checked frame by frame against the Python reference. The exact 250-frame floor needs 1020 bytes of state and about 244 comparisons per frame. Keeping one minimum per block of 10 frames cuts that to 120 bytes and about 24 comparisons, flips 0.02–0.08% of gate decisions, and changes mean recall by at most 0.0002 and modelled battery life by at most 0.07 h (of about 82–90 h) on synthetic audio, 20 seeds × 4 rooms. Limitations: synthetic audio only; host gcc struct sizes and comparison counts, not measured cycles on an ARM core; battery hours inherit the ambient-wearable model.

A figure from its own output

Paper, code and results

• README and method
• Experiments and results
• Code
• Tests

More in Ambient AI assistant

• How much pre-roll buffer does an energy-gated audio pendant need to stop clipping the first word, and what does it cost in RAM and battery?How much pre-roll buffer does an energy-gated audio pendant need to stop clipping the first word? Synthetic audio: no gain in quiet rooms, in the loudest room onset recall 0.866 → 0.980 → 0.999 at 0, 5 and 25 buffered frames (0, 3.1, 15.6 KB) for 90 → 81 → 70 modelled hours; a too-deaf gate is not rescued (stylised)
• Does a fixed energy-gate threshold survive different rooms on an always-on audio pendant, and what does an adaptive one buy in battery life?Does a fixed energy-gate threshold survive different rooms on an always-on audio pendant? Synthetic audio, 20 seeds x 4 background levels: any fixed threshold that keeps speech recall >= 0.9 leaves the gate open all the time in the loudest room (57.6 h, no gating gain), while one adaptive threshold (4x a 5 s running-minimum floor) gives 82-90 h at recall 0.89-1.00 in all four; stylised, nothing measured on hardware
• What limits the battery life of a pendant-style always-on audio wearable?What limits battery life in a pendant-style always-on audio wearable: stylised nRF52840-class power model (mic floor ~170 h; speech-gated ~105 h vs 59 h always streaming) plus a 74-byte VAD gate that compiles for Cortex-M4; no hardware measured
• How much latency and memory does an open-model voice pipeline cost on a CPU, and what reply time does it leave?CPU voice pipeline (Silero VAD, Whisper tiny.en, Qwen2.5-0.5B, Piper) on synthetic speech: per-stage latency/RTF/memory and a reply-latency budget of 0.9–1.5 s at 200–800 ms endpoint hangover, ~0.7 s of it compute. Server CPU, not a phone.
• When should an always-on assistant speak, and does a policy tuned in a simulator keep its annoyance budget in a real room?When an always-on assistant should speak: wake, keyword, threshold and learned policies at equal annoyance, and what breaks when the simulator's room differs from the real one
• How much does speech-recognition error hurt an always-on assistant's trigger policies at equal annoyance?How much speech-recognition error hurts an always-on assistant's trigger policies at equal annoyance: from word error rate 0 to 0.5 a two-word wake phrase loses 74% of its recall and a transcript-reading threshold 76%, an acoustic-only score about 54%, and the transcript policy falls behind the acoustic one at WER 0.3 to 0.4 if its separation shrinks as (1-WER)^2 (stylised).

All 46 Ambient AI assistant projects →

Other research topics

Proper scoring & elicitation28Effort & peer prediction9Verification games & slashing19Detection & tolerances14Markets as routers13Prediction-market design9Paying contributors8Decentralised training systems23Digital twins & sim-to-real164LLM chip design loops40Robot manufacturing loops53Ambient AI assistant46Self-replicating manufacturing34Simulation-first design64Engineered clothing45Task-specific AI accelerators40Algorithmic chip-design loops42Open-source purpose-built robots44The software-engineering loop43Adversarial human imitation41Open-source residential construction41Autonomous multi-angle capture34Sensing, imaging and 3D31Energy supply and dense energy storage39Robot societies among people31Agentic co-design of low-cost robots27Laundry-hanging robots25Machine learning on qualitative data23Robots on the building site27Civilizations & group psychology1History & human sciences1

Produced with AI assistance (Claude). Results are from the code in this repository and are simulated or stylised unless the abstract says otherwise; nothing here implies affiliation with or endorsement by any lab, researcher or company cited. MIT licensed.