File size: 1,624 Bytes
83bddac daa02e7 83bddac | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 | ---
license: other
license_name: campplus-upstream
license_link: https://github.com/modelscope/FunASR
language: [zh]
library_name: coreml
tags: [coreml, ane, speaker-verification, speaker-diarization, campplus, funasr, fluidaudio]
pipeline_tag: audio-classification
---
# CAM++ — CoreML (Apple Neural Engine)
CoreML conversion of FunASR's **CAM++** speaker-embedding model (~7.2M params), for
on-device speaker verification / diarization on Apple Silicon. Upstream:
[iic/speech_campplus_sv_zh-cn_16k-common](https://www.modelscope.cn/models/iic/speech_campplus_sv_zh-cn_16k-common).
## Files
| File | Precision | Compute unit | Role |
|------|-----------|--------------|------|
| `CamPlusPreprocessor.mlmodelc` | FP32 | CPU | waveform → 80-d fbank features |
| `CamPlusPlus.mlmodelc` | FP16 | ANE | fbank → 192-d speaker embedding |
## Pipeline
```
waveform → [Preprocessor fp32/CPU] → fbank [1,T,80]
→ [CAM++ fp16/ANE] → embedding [1,192] (L2-normalize, then cosine for verification/clustering)
```
CAM++ normalizes the fbank internally. The 192-d embedding is used with cosine
similarity for speaker verification and diarization clustering.
## Benchmark — AISHELL-1 speaker verification
| Metric | Value |
|--------|-------|
| **EER** | **0.48%** (20 speakers, 6000 same / 6000 diff trials) |
| same-speaker cosine | 0.805 |
| different-speaker cosine | 0.256 |
AISHELL-1 (clean read Mandarin) is easier than the official CN-Celeb (~6-7%). CoreML↔torch embedding cosine 0.9997-0.99999.
## License
Weights derive from FunASR's CAM++; upstream license applies. Format conversion only.
|