# @stem_antics on Instagram

- **Type:** Video
- **Original URL:** https://www.instagram.com/p/DZbZievgSIU
- **Gondola URL:** https://gondola.cc/posts/66591650-stem-antics-instagram
- **Thumbnail:** https://img.gondola.cc/tr:w-,h-,fo-auto/postThumbnails/54b89ad361.jpg
- **Posted:** 2026-06-11T02:04:07.000+00:00
- **Account Owner:** Stem Antics (@stem_antics) — https://gondola.cc/stem_antics

## Caption

Continuously listening for a wake word on the main A-series application processor would drain the battery in hours. So Apple splits the job across two chips with wildly different power budgets. On iPhone 6S and later, the Always On Processor (AOP) — the same low-power embedded motion coprocessor that counts your steps — taps directly into the microphone signal and runs a tiny acoustic model day and night.

That first-pass detector is a deliberately small deep neural network: 5 layers of just 32 hidden units. It slices incoming audio into roughly 0.01 second frames, converts each into a spectral feature vector, and outputs a probability distribution over speech sounds. An HMM decoder stitches those frame-by-frame predictions together with dynamic programming into a single confidence score for the phrase.

The clever part is what it does NOT do. It does not understand language. It does not transcribe. It answers one question cheaply: did that sound like “Hey Siri”? When the score crosses a lower threshold, the AOP wakes the main processor, which re-runs the buffered audio through a much larger network — 5 layers of 192 hidden units — to confirm before committing real power to your request.

This tiered cheap-gate, expensive-verifier pattern lives in nearly all always-listening hardware:

	•	Alexa and Google Assistant run wake-word spotters on dedicated low-power DSPs, escalating to the main SoC only on a hit
	•	AirPods and watches shrink the model further to survive on a coin-sized battery
	•	Smart speakers hold a rolling buffer so no audio leaves the device until the trigger actually fires

Always on, almost never awake. That gap is the entire design.

#stemantics #embeddedsystems #machinelearning #siri #lowpower

## Stats

- **Views:** 12,807
- **Likes:** 487
- **Shares:** 0
- **Comments:** 13

## Tags

siri, lowpower, embeddedsystems, machinelearning, stemantics

---
Copyright (c) Gondola