Finished LAS Paper

6th-Semester-Spring-2024/DSP/Labs/FinalProject/adding_noise.py

@@ -1,10 +1,15 @@
 import numpy as np
 import matplotlib.pyplot as plt
+
 import scipy.signal
 from scipy.io import wavfile
+import scipy.stats as st
+
 import sounddevice as sd
-import random
 from pesq import pesq
+from pns.noise_suppressor import NoiseSuppressor
+
+import random
 import json
 
 
@@ -38,6 +43,7 @@ def load_audiofile(path):
         sample_rate, sound_data = wavfile.read(path)
         # Make sure it is nparray of floats (trust me bro, normalizing yells at you if its ints)
         sound_data = np.array(sound_data, dtype=np.float64)
+    sound_data = normalize_signal(sound_data)
     return sample_rate, sound_data
 
 
@@ -61,60 +67,125 @@ def add_noise(signal, noise, snr):
     noise_crop *= ratio
 
     noisy_signal = signal + noise_crop
-    return noisy_signal
+    return normalize_signal(noisy_signal)
 
 
-def main():
-    noise_paths = ("noisefiles/white.dat", "noisefiles/train.dat", "noisefiles/street.dat", "noisefiles/exhibition.dat")
+def load_all_signals():
+    signal_sample_rate = None
+    all_signals = []
 
     # Compose signal paths for the 30 sentences
     signal_paths = []
     for i in range(1,30+1):
         signal_paths.append(f"speechfiles/sp{i:02}.wav")
 
-    # SNR in dB
-    snrs = (0, 10, 20, 30)
+    # Load all signals
+    for signal_path in signal_paths:
+        new_signal_sample_rate, signal_data = load_audiofile(signal_path)
+        assert signal_sample_rate == new_signal_sample_rate or signal_sample_rate == None, "Non-uniform signal sampling rates"
+        signal_sample_rate = new_signal_sample_rate
+        all_signals.append(signal_data)
 
-    pesqs = {"unfiltered": {}, "filtered": {}}
+    return signal_sample_rate, all_signals
+
+
+def load_all_noises():
+    noise_paths = ("noisefiles/white.dat", "noisefiles/train.dat", "noisefiles/street.dat", "noisefiles/exhibition.dat")
+    noise_sample_rate = None
+    all_noises = []
+
+    # Load all noises
+    for noise_path in noise_paths:
+        new_noise_sample_rate, noise_data = load_audiofile(noise_path)
+        assert noise_sample_rate == new_noise_sample_rate or noise_sample_rate == None, "Non-uniform noise sampling rates"
+        noise_sample_rate = new_noise_sample_rate
+        all_noises.append(noise_data)
+
+    return noise_sample_rate, all_noises
+
+
+def enhanced(signal, sample_rate):
+    '''
+    noise_suppressor = NoiseSuppressor(sample_rate)
+    frame_size = noise_suppressor.get_frame_size()
+    filtered_signal = np.zeros(len(signal))
+    
+    k = 0
+    while k + frame_size < len(signal):
+        frame = signal[k : k+frame_size]
+        filtered_signal[k : k+frame_size] = noise_suppressor.process_frame(frame)
+        k += frame_size
+    '''
+    filtered_signal = scipy.signal.wiener(signal)
+    return normalize_signal(filtered_signal)
+
+
+def calculate_pesqs(all_signals, all_noises, sample_rate, snrs):
+    noises = ('white', 'train', 'street', 'exhibition')
+    noisy_pesqs = {}
+    filtered_pesqs = {}
 
     for snr in snrs:
-        pesqs["unfiltered"][snr] = {}
-        pesqs["filtered"][snr] = {}
-
-        for noise_path in noise_paths:
-            pesqs["filtered"][snr][noise_path[:-4]] = []
-            pesqs["unfiltered"][snr][noise_path[:-4]] = []
-
-            for signal_path in signal_paths:
-                noise_sample_rate, noise_data = load_audiofile(noise_path)
-                signal_sample_rate, signal_data = load_audiofile(signal_path)
-
-                assert signal_sample_rate == noise_sample_rate, "Signal and noise sampling rates didn't match."
-                sample_rate = signal_sample_rate
-
+        noisy_pesqs[snr] = []
+        filtered_pesqs[snr] = []
+        for i, noise_data in enumerate(all_noises):
+            for j, signal_data in enumerate(all_signals):
                 noisy_signal = add_noise(signal_data, noise_data, snr)
-                filtered_signal = scipy.signal.wiener(noisy_signal)
+                filtered_signal = enhanced(noisy_signal, sample_rate)
+                #wavfile.write(f"noisy/sp{j+1:02}_{noises[i]}_snr{snr}.wav", sample_rate, )
+                #sd.play(normalize_signal(signal_data), samplerate=sample_rate, blocking=True)
+                #sd.play(normalize_signal(noisy_signal), samplerate=sample_rate, blocking=True)
+                #sd.play(normalize_signal(filtered_signal), samplerate=sample_rate, blocking=True)
             
                 noisy_pesq = pesq(sample_rate, signal_data, noisy_signal, mode='nb')
                 filtered_pesq = pesq(sample_rate, signal_data, filtered_signal, mode='nb')
 
-                pesqs["filtered"][snr][noise_path[:-4]].append(noisy_pesq)
-                pesqs["unfiltered"][snr][noise_path[:-4]].append(filtered_pesq)
+                noisy_pesqs[snr].append(noisy_pesq)
+                filtered_pesqs[snr].append(filtered_pesq)
 
-    pesqs_json = json.dumps(pesqs, indent=4)
-    with open("pesqs.json", "w") as outfile:
-        outfile.write(pesqs_json)
+    return noisy_pesqs, filtered_pesqs
 
-    plt.plot(snrs, noisy_pesqs, label="PESQ Unfiltered")
-    plt.plot(snrs, filtered_pesqs, label="PESQ Filtered")
+
+def main():
+    # SNR in dB
+    snrs = (0, 10, 20, 30)
+
+    # Load all signals and all noises
+    signal_sample_rate, all_signals = load_all_signals()
+    noise_sample_rate, all_noises = load_all_noises()
+
+    assert signal_sample_rate == noise_sample_rate, "Signal and noise sampling rates didn't match."
+    sample_rate = signal_sample_rate
+
+    noisy_pesqs, filtered_pesqs = calculate_pesqs(all_signals, all_noises, sample_rate, snrs)
+
+    n_pesq = [noisy_pesqs[snr][0] for snr in snrs]
+    f_pesq = [filtered_pesqs[snr][0] for snr in snrs]
+    plt.plot(snrs, n_pesq, label="Noisy")
+    plt.plot(snrs, f_pesq, label="Filtered")
     plt.xlabel("SNR [dB]")
     plt.ylabel("PESQ")
     plt.legend()
     plt.show()
     
-    #sd.play(normalize_signal(noisy_signal), samplerate=sample_rate, blocking=True)
-    #sd.play(normalize_signal(filtered_signal), samplerate=sample_rate, blocking=True)
+    fig, axes = plt.subplots(nrows=1, ncols=4, figsize=(9,9), sharey=True)
 
+    # Calculate the 95% confidence interval for the noise and filter PESQ at each SNR
+    for i, snr in enumerate(snrs):
+        data_noise = np.array(noisy_pesqs[snr])
+        data_filter = np.array(filtered_pesqs[snr])
+
+        noise_ci = st.t.interval(0.95, len(data_noise)-1, loc=np.mean(data_noise), scale=st.sem(data_noise))
+        filter_ci = st.t.interval(0.95, len(data_filter)-1, loc=np.mean(data_filter), scale=st.sem(data_filter))
+        
+        axes[i].vlines((1,2), (min(noise_ci), min(filter_ci)), (max(noise_ci), max(filter_ci)))
+        axes[i].set_xticks(np.arange(1, 3), labels=["Unfiltered", "Filtered"])
+        axes[i].set_xlim(0.25, 2.75)
+        axes[i].set_xlabel(f"{snr}[dB] SNR")
+    axes[0].set_ylabel("PESQ")
+
+    #plt.savefig("PESQ_Confidence_Interval.png")
+    plt.show()
 
 if __name__ == "__main__":
     main()