cronekorkn/bootshorn_investigation
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

wip

Browse source
This commit is contained in:
CroneKorkN 2025-05-30 21:31:16 +02:00
parent abee103ed9
commit 788416adb6
Signed by: cronekorkn
SSH key fingerprint: SHA256:v0410ZKfuO1QHdgKBsdQNF64xmTxOF8osF1LIqwTcVw
1 changed files with 43 additions and 38 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

81
process_chunks.py
View file

@ -7,15 +7,15 @@ import matplotlib.pyplot as plt
import soundfile as sf
import scipy.signal
from scipy.fft import fft, fftfreq
from datetime import datetime
import shutil
INPUT_DIR = "chunks"
OUTPUT_DIR = "chunks/processed"
CHUNK_SECONDS = 1
TOLERANCE = 1
CHUNK_DIR = "chunks"
PROCESSED_CHUNK_DIR = "chunks/processed"
EVENT_DIR = "events"
SAMPLE_SECONDS = 1
TOLERANCE = 2
OVERTONE_TOLERANCE = TOLERANCE * 2
THRESHOLD_BASE = 0.5
THRESHOLD_BASE = 0.
THRESHOLD_OCT = THRESHOLD_BASE / 10
CLIP_PADDING_BEFORE = 1
CLIP_PADDING_AFTER = 6
@ -24,25 +24,8 @@ OVERTONE_FREQ = TARGET_FREQ * 2
NFFT = 32768
SKIP_SECONDS = 10
def detect_event(chunk, samplerate):
freqs, times, Sxx = scipy.signal.spectrogram(chunk, samplerate, nperseg=NFFT)
idx_base = np.where((freqs >= TARGET_FREQ - TOLERANCE) & (freqs <= TARGET_FREQ + TOLERANCE))[0]
idx_oct = np.where((freqs >= OVERTONE_FREQ - OVERTONE_TOLERANCE) & (freqs <= OVERTONE_FREQ + OVERTONE_TOLERANCE))[0]
if len(idx_base) == 0 or len(idx_oct) == 0:
return False
base_energy = np.mean(Sxx[idx_base])
oct_energy = np.mean(Sxx[idx_oct])
total_energy = np.mean(Sxx, axis=0).max()
fft_vals = np.abs(fft(chunk))
freqs = fftfreq(len(chunk), 1/samplerate)
peak_freq = freqs[np.argmax(fft_vals)]
is_peak_near_target = TARGET_FREQ - TOLERANCE <= peak_freq <= TARGET_FREQ + TOLERANCE
return is_peak_near_target and base_energy > THRESHOLD_BASE * total_energy and oct_energy > THRESHOLD_OCT * total_energy
def process_chunk(filename):
input_path = os.path.join(INPUT_DIR, filename)
input_path = os.path.join(CHUNK_DIR, filename)
print(f"🔍 Verarbeite {input_path}...")
# Frequenzanalyse und Event-Erkennung
@ -50,29 +33,47 @@ def process_chunk(filename):
if data.ndim > 1:
data = data[:, 0] # nur Kanal 1
chunk_samples = int(CHUNK_SECONDS * samplerate)
chunk_samples = int(SAMPLE_SECONDS * samplerate)
skip_samples = int(SKIP_SECONDS * samplerate)
padding_before = int(CLIP_PADDING_BEFORE * samplerate)
padding_after = int(CLIP_PADDING_AFTER * samplerate)
chunk_start_str = os.path.splitext(filename)[0]
chunk_start_dt = datetime.datetime.strptime(chunk_start_str, "%Y%m%d-%H%M%S")
i = 0
last_event = -skip_samples
while i + chunk_samples <= len(data):
chunk = data[i:i+chunk_samples]
if i - last_event >= skip_samples and detect_event(chunk, samplerate):
clip = data[i:i+chunk_samples]
freqs, times, Sxx = scipy.signal.spectrogram(clip, samplerate, nperseg=NFFT)
idx_base = np.where((freqs >= TARGET_FREQ - TOLERANCE) & (freqs <= TARGET_FREQ + TOLERANCE))[0]
idx_oct = np.where((freqs >= OVERTONE_FREQ - OVERTONE_TOLERANCE) & (freqs <= OVERTONE_FREQ + OVERTONE_TOLERANCE))[0]
if len(idx_base) == 0 or len(idx_oct) == 0:
return False
base_energy = np.mean(Sxx[idx_base])
oct_energy = np.mean(Sxx[idx_oct])
total_energy = np.mean(Sxx, axis=0).max()
fft_vals = np.abs(fft(clip))
freqs = fftfreq(len(clip), 1/samplerate)
peak_freq = freqs[np.argmax(fft_vals)]
is_peak_near_target = TARGET_FREQ - TOLERANCE <= peak_freq <= TARGET_FREQ + TOLERANCE
event_detected = is_peak_near_target and base_energy > THRESHOLD_BASE * total_energy and oct_energy > THRESHOLD_OCT * total_energy
if i - last_event >= skip_samples and event_detected:
clip_start = max(0, i - padding_before)
clip_end = min(len(data), i + chunk_samples + padding_after)
clip = data[clip_start:clip_end]
chunk_start_str = os.path.splitext(filename)[0]
chunk_start_dt = datetime.strptime(chunk_start_str, "%Y%m%d-%H%M%S")
event_offset = (i - padding_before) / samplerate
event_time_dt = chunk_start_dt + datetime.timedelta(seconds=event_offset)
event_time = event_time_dt.strftime("%Y%m%d-%H%M%S")
base_name = os.path.splitext(filename)[0]
flac_out = os.path.join(OUTPUT_DIR, f"{base_name}_{event_time}.flac")
png_out = os.path.join(OUTPUT_DIR, f"{base_name}_{event_time}.png")
flac_out = os.path.join(EVENT_DIR, f"{base_name}_{event_time}.flac")
png_out = os.path.join(EVENT_DIR, f"{base_name}_{event_time}.png")
sf.write(flac_out, clip, samplerate, format='FLAC')
plt.figure()
@ -84,24 +85,28 @@ def process_chunk(filename):
plt.savefig(png_out)
plt.close()
print(f"🎯 Ereignis erkannt bei {event_time}, gespeichert: {flac_out}, {png_out}")
print(f"Event: {event_time} peak_freq: {int(peak_freq)} base_energy: {int(base_energy)} oct_energy: {int(oct_energy)} total_energy: {int(total_energy)}")
last_event = i
i += skip_samples
else:
i += chunk_samples
# Datei verschieben
output_path = os.path.join(OUTPUT_DIR, filename)
output_path = os.path.join(PROCESSED_CHUNK_DIR, filename)
#shutil.move(input_path, output_path)
print(f"✅ Verschoben nach {output_path}")
def main():
os.makedirs(OUTPUT_DIR, exist_ok=True)
os.makedirs(EVENT_DIR, exist_ok=True)
os.makedirs(PROCESSED_CHUNK_DIR, exist_ok=True)
with concurrent.futures.ProcessPoolExecutor() as executor:
files = [f for f in os.listdir(INPUT_DIR) if f.endswith(".flac")]
executor.map(process_chunk, files)
for file in os.listdir(CHUNK_DIR):
if file.endswith(".flac"):
process_chunk(file)
# with concurrent.futures.ProcessPoolExecutor() as executor:
# files = [f for f in os.listdir(CHUNK_DIR) if f.endswith(".flac")]
# executor.map(process_chunk, files)
if __name__ == "__main__":
main()