Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition
Paper • 1804.03209 • Published
Datasets:
audio dict | label class label 30
classes |
|---|---|
{
"bytes": [
82,
73,
70,
70,
36,
125,
0,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
100,
97,
116,
... | 14one |
{
"bytes": [
82,
73,
70,
70,
36,
125,
0,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
100,
97,
116,
... | 2call |
{
"bytes": [
82,
73,
70,
70,
36,
125,
0,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
100,
97,
116,
... | 3close |
{
"bytes": [
82,
73,
70,
70,
36,
125,
0,
0,
87,
65,
86,
69,
102,
109,
116,
32,
16,
0,
0,
0,
1,
0,
1,
0,
128,
62,
0,
0,
0,
125,
0,
0,
2,
0,
16,
0,
100,
97,
116,
... | 2call |
{"bytes":"UklGRiR9AABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQB9AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA(...TRUNCATED) | 11next |
{"bytes":"UklGRiR9AABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQB9AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA(...TRUNCATED) | 18record |
{"bytes":"UklGRiR9AABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQB9AABWADIANABrAGQAWABpAI8AcgBRAHIAeg(...TRUNCATED) | 22six |
{"bytes":"UklGRiR9AABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQB9AAAq/Tv7tPjj+p38Tf16+7b5EfkV+Vr6Sv(...TRUNCATED) | 21seven |
{"bytes":"UklGRiR9AABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQB9AAAGAAEAGgAUAPH/4v/h//b/EQAKAPv/CQ(...TRUNCATED) | 19right |
{"bytes":"UklGRiR9AABXQVZFZm10IBAAAAABAAEAgD4AAAB9AAACABAAZGF0YQB9AAD3//7/9f/q/9z/4f/z//P/9//3/+7/8/(...TRUNCATED) | 22six |
End of preview. Expand in Data Studio
KWS Smart-Device Commands (30-class)
Merged, length-normalized keyword-spotting dataset for on-device voice control (Nordic nRF5340 + Edge Impulse). Every clip is 16 kHz mono, fixed 1 s, RMS-leveled and silence-padded. 1,000 clips per class. Real CC-BY audio only — no TTS.
- Classes (30): answer, back, call, close, down, eight, five, four, go, help, left, next, nine, no, one, open, play, previous, record, right, search, seven, six, start, stop, three, two, up, yes, zero
- Features:
audio(WAV bytes + path),label(ClassLabel) - Splits: stratified train/test (80/20)
- License: CC-BY 4.0
Usage
from datasets import load_dataset, Audio
import soundfile as sf
import io
ds = load_dataset("snowballlab/30-keywords-v2")
print(ds)
print(ds["train"].features["label"].names)
# Decode one clip with soundfile (no torch needed):
ex = ds["train"][0]
y, sr = sf.read(io.BytesIO(ex["audio"]["bytes"]))
print(y.shape, sr, ds["train"].features["label"].int2str(ex["label"]))
# Or, if you have torchcodec installed:
# ds = ds.cast_column("audio", Audio(sampling_rate=16000))
Sources & attribution
- Google Speech Commands v0.02 — CC-BY 4.0 (Warden, 2018, arXiv:1804.03209), including
up,down, andgo - Multilingual Spoken Words Corpus (English) — CC-BY 4.0 (MLCommons)
Preprocessing matches the MLPerf Tiny KWS reference (16 kHz, MFCC 30/20 ms, 40 mel, 10 coeff, 20–4000 Hz).
- Downloads last month
- 21