python通过librosa读取一段wav语音信号，并对读取的语音信号分别进行普通滤波、维纳滤波、卡尔曼滤波、自适应滤波等滤波方法消除或压制嗓声并画出他们的基音频率曲线展示信号的代码

好的，以下是代码实现：

import librosa
import numpy as np
import matplotlib.pyplot as plt
from scipy.signal import wiener, savgol_filter, medfilt, lfilter, butter, freqz

# 读取wav语音信号
y, sr = librosa.load('speech.wav', sr=16000)

# 计算基音频率曲线
pitches, magnitudes = librosa.piptrack(y=y, sr=sr)
pitch = np.nanmean(librosa.hz_to_midi(pitches[magnitudes > np.median(magnitudes)]) - 69)

# 普通滤波
y_filt1 = savgol_filter(y, window_length=31, polyorder=3)

# 维纳滤波
y_filt2 = wiener(y)

# 中值滤波
y_filt3 = medfilt(y, kernel_size=11)

# 卡尔曼滤波
def kalman_filter(z_meas, x_esti, P):
    # 状态转移矩阵
    F = np.array([[1, 1], [0, 1]])
    # 系统噪声方差
    Q = np.array([[0.05, 0.05], [0.05, 0.2]])
    # 观测矩阵
    H = np.array([1, 0]).reshape(1, 2)
    # 观测噪声方差
    R = np.array([0.2]).reshape(1, 1)
    # 状态预测
    x_predict = np.dot(F, x_esti)
    P_predict = np.dot(F, np.dot(P, F.T)) + Q
    # 计算卡尔曼增益
    K = np.dot(P_predict, np.dot(H.T, np.linalg.inv(np.dot(H, np.dot(P_predict, H.T)) + R)))
    # 更新状态估计值和误差协方差矩阵
    x_estimate = x_predict + np.dot(K, (z_meas - np.dot(H, x_predict)))
    P_estimate = np.dot((np.eye(2) - np.dot(K, H)), P_predict)
    return x_estimate, P_estimate

x_esti = np.array([0, 0]).reshape(2, 1)
P = 1.0 * np.eye(2)
y_filt4 = np.zeros_like(y)
for i in range(len(y)):
    if np.isnan(pitches[i]):
        pitch = pitch
    else:
        pitch = librosa.hz_to_midi(pitches[i]) - 69
    z_meas = np.array([pitch, 0]).reshape(2, 1)
    x_esti, P = kalman_filter(z_meas, x_esti, P)
    y_filt4[i] = y[i] - x_esti[0]

# 自适应滤波
def adaptive_filter(input_sig, filter_order, mu, delta):
    # 初始化
    w = np.zeros(filter_order)
    y = np.zeros_like(input_sig)
    e = np.zeros_like(input_sig)
    for n in range(filter_order, len(input_sig)):
        # 构造输入向量
        u = np.flip(input_sig[n-filter_order:n], axis=0)
        # 计算预测信号
        y[n] = np.dot(w, u)
        # 计算误差信号
        e[n] = input_sig[n] - y[n]
        # 更新滤波系数
        w = w + 2 * mu * e[n] * u / (delta + np.dot(u, u))
    return e

y_filt5 = adaptive_filter(y, filter_order=50, mu=0.005, delta=1)

# 绘制基音频率曲线和滤波后的信号
plt.figure(figsize=(12, 8))
plt.subplot(3, 2, 1)
plt.plot(pitches)
plt.title('Pitch')
plt.subplot(3, 2, 2)
plt.plot(y)
plt.title('Original Signal')
plt.subplot(3, 2, 3)
plt.plot(y_filt1)
plt.title('Savitzky-Golay Filter')
plt.subplot(3, 2, 4)
plt.plot(y_filt2)
plt.title('Wiener Filter')
plt.subplot(3, 2, 5)
plt.plot(y_filt3)
plt.title('Median Filter')
plt.subplot(3, 2, 6)
plt.plot(y_filt4)
plt.title('Kalman Filter')
plt.tight_layout()
plt.show()

其中，普通滤波使用了Savitzky-Golay滤波器；维纳滤波使用了Scipy库中的wiener函数；中值滤波使用了Scipy库中的medfilt函数；自适应滤波使用了LMS自适应滤波算法。另外，卡尔曼滤波的代码实现使用了Python中的numpy库。