import numpy as np
import matplotlib.pyplot as plt
import librosa
from sklearn.decomposition import TruncatedSVD
from scipy.linalg import svd, polar
import pandas as pd

def load_signal_from_csv(csv_file_path):
    csv_data = pd.read_csv(csv_file_path, header=None, names=['Values'])
    y = csv_data['Values'].to_numpy()

    return y

def create_spectrogram(y, n_fft=1024, hop_length=2, win_length=128):
    stft = librosa.stft(y, n_fft=n_fft, hop_length=hop_length, win_length=win_length)
    spectrogram = np.abs(stft)

    stft_real = stft.real
    spectrogram_db = librosa.amplitude_to_db(spectrogram, ref=np.max)

    N = spectrogram_db.shape[0]

    # Visualize the spectrogram
    fig, ax = plt.subplots(figsize=(10, 5))
    img = librosa.display.specshow(spectrogram_db,
                                   x_axis='time',
                                   y_axis='hz',
                                   fmin=0,
                                   fmax=17000,
                                   ax=ax)
    ax.set_title('Spectrogram', fontsize=20)
    fig.colorbar(img, ax=ax, format='%0.002f')
    plt.show()

    return spectrogram, stft_real

def create_pca(stft_real):
    covX = np.dot(stft_real.T, stft_real)
    d, V = np.linalg.eig(covX)
    Xpca = np.real(np.dot(stft_real, V[:, :2]))

    fig, axs = plt.subplots(2, 1, figsize=(12, 10))
    axs[0].semilogy(np.abs(d), '.')
    axs[0].grid()

    axs[1].plot(Xpca[:, 0], Xpca[:, 1], '.k')
    axs[1].grid()

    # Create a new figure for PCA results
    plt.figure(figsize=(12, 5))
    plt.plot(Xpca[:, 0], Xpca[:, 1], '.k')
    plt.axhline(y=2, color='r', linestyle='--', xmin=0, xmax=1)
    plt.axhline(y=-2, color='r', linestyle='--', xmin=0, xmax=1)
    plt.title("PCA Lehce klepe s hranicemi")
    plt.grid()
    plt.savefig('PCA Clustering Light Knocking (Separate)')
    plt.show()

    plt.show()

    return Xpca

def create_svd(stft_real):
    svd_model = TruncatedSVD(n_components=2, n_iter=7, random_state=50)
    Xsvd = svd_model.fit_transform(stft_real)

    plt.figure(figsize=(12, 5))
    plt.plot(Xsvd[:, 0], Xsvd[:, 1], '.k')
    plt.title("SVD Results")
    plt.axhline(y=2, color='r', linestyle='--', xmin=0, xmax=1)
    plt.axhline(y=-2, color='r', linestyle='--', xmin=0, xmax=1)
    plt.grid()
    plt.show()

    return Xsvd

def Borders_PCA_SVD(Xpca, Xsvd, pca_upper_bound=2, pca_lower_bound=-2, svd_upper_bound=2, svd_lower_bound=-2):
    fig, axs = plt.subplots(1, 2, figsize=(18, 5))

    # Plot PCA with borders
    axs[0].axhline(y=pca_upper_bound, color='r', linestyle='--', xmin=0, xmax=1)
    axs[0].axhline(y=pca_lower_bound, color='r', linestyle='--', xmin=0, xmax=1)
    axs[0].plot(Xpca[:, 0], Xpca[:, 1], '.k')
    axs[0].grid()
    axs[0].set_title("PCA Results with Borders")

    # Plot SVD with borders
    axs[1].axhline(y=svd_upper_bound, color='r', linestyle='--', xmin=0, xmax=1)
    axs[1].axhline(y=svd_lower_bound, color='r', linestyle='--', xmin=0, xmax=1)
    axs[1].plot(Xsvd[:, 0], Xsvd[:, 1], '.k')
    axs[1].grid()
    axs[1].set_title("SVD Results with Borders")

    plt.show()

# Load signal from CSV
csv_file_path = # Nacist csv file
y = load_signal_from_csv(csv_file_path)

# Create and plot spectrogram
spectrogram, stft_real = create_spectrogram(y)

# Apply PCA and plot results
Xpca = create_pca(stft_real)

# Apply SVD and plot results
Xsvd = create_svd(stft_real)

# Plot PCA and SVD with borders
Borders_PCA_SVD(Xpca, Xsvd, pca_upper_bound=2, pca_lower_bound=-2, svd_upper_bound=2, svd_lower_bound=-2)