File size: 2,303 Bytes
5a7104f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
48
49
50
51
52
53
54
55
"""Next Signal: self-supervised representations for inertial sensor windows (UCI HAR).

Shared by the notebook, inference.py and (ported) web/core.js:
* normalise: per-channel standardisation with statistics from the pre-training subjects
* encode: frozen 1-D CNN encoder exported to nlrt (BatchNorm folded) -> 64-d embedding
* stats_features: 54 hand-made features (6 per channel)
* probe: standardise + multinomial logistic regression (weights fitted in the notebook)
* perturb: deterministic noise / amplitude / circular time-shift used by the Space
"""
import numpy as np

from nlrt import Mulberry32, forward

ACTIVITIES = ["walking", "walking upstairs", "walking downstairs", "sitting", "standing", "lying"]
CHANNELS = ["body_acc_x", "body_acc_y", "body_acc_z", "body_gyro_x", "body_gyro_y", "body_gyro_z",
            "total_acc_x", "total_acc_y", "total_acc_z"]


def normalise(x, mean, std):
    return (np.asarray(x, dtype=np.float64) - np.asarray(mean)[:, None]) / np.asarray(std)[:, None]


def encode(layers, xn):
    """xn: (N, 9, 128) normalised windows -> (N, 64)."""
    return forward(layers, xn)


def stats_features(x):
    """x: (N, 9, 128) raw windows -> (N, 54): mean, std, min, max, mean |diff|, mean square per channel."""
    x = np.asarray(x, dtype=np.float64)
    mean = x.mean(axis=2)
    std = np.sqrt(((x - mean[:, :, None]) ** 2).mean(axis=2))
    feats = [mean, std, x.min(axis=2), x.max(axis=2), np.abs(np.diff(x, axis=2)).mean(axis=2), (x * x).mean(axis=2)]
    return np.stack(feats, axis=2).reshape(len(x), -1)


def probe_proba(probe, Z):
    Zs = (np.asarray(Z) - probe["mu"]) / probe["sd"]
    z = Zs @ np.asarray(probe["W"]).T + probe["b"]
    z = z - z.max(axis=1, keepdims=True)
    e = np.exp(z)
    return e / e.sum(axis=1, keepdims=True)


def perturb(x, noise=0.0, scale=1.0, shift=0, seed=0):
    """x: (9, 128) raw window. Adds Gaussian noise (std = noise x channel std), scales amplitude, shifts in time."""
    x = np.asarray(x, dtype=np.float64)
    rng = Mulberry32(seed)
    sd = np.sqrt(((x - x.mean(axis=1, keepdims=True)) ** 2).mean(axis=1))
    out = np.empty_like(x)
    for c in range(x.shape[0]):
        for t in range(x.shape[1]):
            out[c, t] = x[c, (t - shift) % x.shape[1]] * scale + noise * sd[c] * rng.normal()
    return out