neuralworm commited on
Commit
c03af22
·
verified ·
1 Parent(s): 59735a8

Update cognitive_mapping_probe/signal_analysis.py

Browse files
cognitive_mapping_probe/signal_analysis.py CHANGED
@@ -1,56 +1,55 @@
1
  import numpy as np
2
  from scipy.fft import rfft, rfftfreq
3
- from scipy.signal import hilbert
4
- from typing import Dict, Tuple
5
 
6
  def analyze_cognitive_signal(
7
  state_deltas: np.ndarray,
8
- sampling_rate: float = 1.0 # Wir nehmen an, dass jeder "Step" eine Zeiteinheit ist
9
- ) -> Dict[str, float]:
10
  """
11
  Führt eine fortgeschrittene Signalverarbeitungs-Analyse auf der Zeitreihe der
12
- State Deltas durch, um den "kognitiven Frequenz-Fingerabdruck" zu extrahieren.
13
  """
14
- if len(state_deltas) < 2:
15
- return {}
 
 
 
16
 
17
- # 1. Fourier-Analyse (FFT)
18
  n = len(state_deltas)
19
- yf = rfft(state_deltas - np.mean(state_deltas)) # Entferne DC-Offset
20
  xf = rfftfreq(n, 1 / sampling_rate)
21
 
22
  power_spectrum = np.abs(yf)**2
23
 
24
- # 2. Extrahiere Metriken aus dem Spektrum
 
 
 
25
  if len(power_spectrum) > 1:
26
- # Dominante Frequenz (ohne die 0-Hz-Komponente)
27
  dominant_freq_index = np.argmax(power_spectrum[1:]) + 1
28
  dominant_frequency = xf[dominant_freq_index]
29
 
30
- # Spektrale Entropie (Maß für die Komplexität/Rauschhaftigkeit)
 
 
 
 
31
  prob_dist = power_spectrum / np.sum(power_spectrum)
32
- prob_dist = prob_dist[prob_dist > 0] # Vermeide log(0)
33
  spectral_entropy = -np.sum(prob_dist * np.log2(prob_dist))
34
- else:
35
- dominant_frequency = 0.0
36
- spectral_entropy = 0.0
37
-
38
- # 3. Hilbert-Transformation zur Phasen-Analyse (hier nur als Beispiel,
39
- # da die Interpretation komplex ist und im UI schwer darstellbar)
40
- # analytic_signal = hilbert(state_deltas)
41
- # instantaneous_phase = np.unwrap(np.angle(analytic_signal))
42
- # instantaneous_frequency = (np.diff(instantaneous_phase) / (2.0*np.pi) * sampling_rate)
43
 
44
  return {
45
- "dominant_frequency": float(dominant_frequency),
46
- "spectral_entropy": float(spectral_entropy),
47
  }
48
 
49
  def get_power_spectrum_for_plotting(state_deltas: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
50
  """
51
  Berechnet das Leistungsspektrum speziell für die Visualisierung.
52
  """
53
- if len(state_deltas) < 2:
54
  return np.array([]), np.array([])
55
 
56
  n = len(state_deltas)
 
1
  import numpy as np
2
  from scipy.fft import rfft, rfftfreq
3
+ from typing import Dict, Optional, Tuple
 
4
 
5
  def analyze_cognitive_signal(
6
  state_deltas: np.ndarray,
7
+ sampling_rate: float = 1.0
8
+ ) -> Dict[str, Optional[float]]:
9
  """
10
  Führt eine fortgeschrittene Signalverarbeitungs-Analyse auf der Zeitreihe der
11
+ State Deltas durch und gibt hardware-unabhängige Metriken zurück.
12
  """
13
+ if len(state_deltas) < 10:
14
+ return {
15
+ "dominant_period_steps": None,
16
+ "spectral_entropy": None,
17
+ }
18
 
 
19
  n = len(state_deltas)
20
+ yf = rfft(state_deltas - np.mean(state_deltas))
21
  xf = rfftfreq(n, 1 / sampling_rate)
22
 
23
  power_spectrum = np.abs(yf)**2
24
 
25
+ dominant_frequency = None
26
+ spectral_entropy = None
27
+ dominant_period_steps = None
28
+
29
  if len(power_spectrum) > 1:
30
+ # Finde dominante Frequenz (ohne 0-Hz)
31
  dominant_freq_index = np.argmax(power_spectrum[1:]) + 1
32
  dominant_frequency = xf[dominant_freq_index]
33
 
34
+ # FINALE KORREKTUR: Berechne die Periodendauer in "Steps"
35
+ if dominant_frequency > 1e-9: # Schutz vor Division durch Null
36
+ dominant_period_steps = 1 / dominant_frequency
37
+
38
+ # Berechne Spektrale Entropie
39
  prob_dist = power_spectrum / np.sum(power_spectrum)
40
+ prob_dist = prob_dist[prob_dist > 1e-12]
41
  spectral_entropy = -np.sum(prob_dist * np.log2(prob_dist))
 
 
 
 
 
 
 
 
 
42
 
43
  return {
44
+ "dominant_period_steps": float(dominant_period_steps) if dominant_period_steps is not None else None,
45
+ "spectral_entropy": float(spectral_entropy) if spectral_entropy is not None else None,
46
  }
47
 
48
  def get_power_spectrum_for_plotting(state_deltas: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
49
  """
50
  Berechnet das Leistungsspektrum speziell für die Visualisierung.
51
  """
52
+ if len(state_deltas) < 10:
53
  return np.array([]), np.array([])
54
 
55
  n = len(state_deltas)