Heiheihaha17's picture
Add files using upload-large-folder tool
6a16e42 verified
Raw History Blame Contribute Delete
2.94 kB
from __future__ import annotations
import hashlib
import json
import math
from pathlib import Path
import numpy as np
SR = 16000
STEP = 1600
VERSION = 'dish_dynamic_demo_v1'
def speech_activity(wave, offset):
"""Frozen 10 ms energy activity labels, including natural phonetic pauses."""
n=len(wave)
energy=np.mean(np.pad(wave,(0,(-n)%160)).reshape(-1,160).astype(float)**2,axis=1)
mask=energy>max(float(energy.max())*.001,1e-10)
spans=[];start=None
for i,on in enumerate(np.r_[mask,False]):
if on and start is None:start=i*160
if not on and start is not None:
spans.append([offset+start,offset+min(i*160,n)]);start=None
return spans
def read(path):
return json.loads(Path(path).read_text())
def write(path, value):
path = Path(path)
path.parent.mkdir(parents=True, exist_ok=True)
path.write_text(json.dumps(value, ensure_ascii=False, indent=2, allow_nan=False) + '\n')
def jsonl(path, values):
path = Path(path)
path.parent.mkdir(parents=True, exist_ok=True)
with path.open('w') as f:
for v in values:
f.write(json.dumps(v, ensure_ascii=False, allow_nan=False) + '\n')
def sha(path):
h = hashlib.sha256()
with Path(path).open('rb') as f:
for b in iter(lambda: f.read(1024 * 1024), b''):
h.update(b)
return h.hexdigest()
def uid(*parts):
return hashlib.sha256('/'.join(map(str, (VERSION,) + parts)).encode()).hexdigest()[:24]
def position(entity, t):
"""Minimum-jerk horizontal displacement; fixed height/yaw, finite acceleration."""
u = np.clip(np.asarray(t, dtype=float) / entity['motion_duration_s'], 0., 1.)
s = 10*u**3 - 15*u**4 + 6*u**5
return np.asarray(entity['initial_xyz_m']) + np.expand_dims(s, -1)*np.asarray(entity['displacement_m'])
def velocity(entity, t):
u = np.clip(np.asarray(t, dtype=float) / entity['motion_duration_s'], 0., 1.)
speed = (30*u**2 - 60*u**3 + 30*u**4) / entity['motion_duration_s']
return np.expand_dims(speed, -1)*np.asarray(entity['displacement_m'])
def geometry(source, robot, t):
d = position(source, t) - position(robot, t)
x, y, z = map(float, d)
az = (-(math.degrees(math.atan2(x, -z))-robot['yaw_deg'])+180) % 360-180
return dict(doa_deg=[az, math.degrees(math.atan2(y, math.hypot(x, z)))],
distance_m=float(np.linalg.norm(d)), xyz_m=position(source, t).tolist())
def active(source, sample):
return any(a <= sample < b for event in source['events'] for a, b in event['activity_samples'])
def angular(a, b):
a, b = np.radians(a), np.radians(b)
return float(np.degrees(np.arccos(np.clip(np.sin(a[1])*np.sin(b[1]) +
np.cos(a[1])*np.cos(b[1])*np.cos(a[0]-b[0]), -1., 1.))))
def grid(scene):
return list(range(0, scene['duration_samples'], STEP))
def tokens(text):
import re
return re.findall(r"[a-z0-9]+(?:'[a-z0-9]+)*", text.lower())