arXiv cs.CL· Daniel Varela S\'anchez, Guillermo Mart\'in-Coello Ju\'arez·· 6 天前AI 评分54
Oído：在 5 美元微控制器上无神经网络加速器运行开放词表语音识别
O\'ido: Open-Vocabulary Speech Recognition on a $5 Microcontroller without a Neural Accelerator
AI 导读
论文提出 Oído，一个完全在 ESP32-S3（240 MHz 双核 Xtensa LX7、512 KB SRAM、8 MB PSRAM、16 MB flash，无神经网络加速器）上运行的开放词表英语语音识别推理引擎，执行 13M 参数 Conformer-CTC 模型，采用 int8 权重、逐帧动态 int8 激活和 int8 注意力。
正文 · 原文
Abstract:Speech recognition on low-cost microcontrollers is, in practice, limited to keyword spotting or a fixed list of commands. We present Oído, an inference engine that runs open-vocabulary English speech recognition entirely on an ESP32-S3 (240 MHz dual-core Xtensa LX7, 512 KB of SRAM, 8 MB of PSRAM, 16 MB of flash, no neural accelerator). Oído executes a 13M-parameter Conformer-CTC model with int8 weights, per-frame dynamic int8 activations and int8 attention on the chip's SIMD instructions, with a schedule designed around flash and PSRAM bandwidth. On LibriSpeech test-clean/test-other the on-device arithmetic reaches 3.70%/8.23% word error rate (WER), within 0.12 points of full precision, and 3.33%/7.15% with an on-chip 1.3M-parameter language model; a transducer head gives 2.99%/6.67% and an int4 variant fits in 8.3 MB. This is 2.3-2.8x fewer errors than the vendor's recognizer on the same chip, and fewer than Whisper this http URL and Moonshine tiny in full precision on a laptop, including in real noise and reverberation. Measured on a physical board, the firmware runs at 1.97x the duration of the audio on one core, with transcripts identical to an instruction-exact emulator: accurate, but not yet real time (a 4 s command is transcribed about 8.7 s after it ends). The two-core schedule gives wrong results on silicon and is disabled. A streaming variant shows partial text while the speaker talks and delivers the final text 4.1-11.1 s after clips of 3.6-10.6 s (7.2-22.4 s in utterance mode) at 4.86%/11.03% WER; a Spanish model makes 34-60% fewer errors than multilingual Whisper tiny. Code and models are released.
| Comments: | 11 pages. Code, models and raw board measurements: this https URL |
| Subjects: | Sound (cs.SD); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS) |
| Cite as: | arXiv:2610.13447 [cs.SD] |
| (or arXiv:2610.13447v1 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2610.13447 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Daniel Varela [view email]
[v1]
Fri, 9 Oct 2026 11:48:37 UTC (31 KB)
来源：arXiv cs.CL · arxiv.org