# -*- coding: utf-8 -*-
# 读Wave文件并且绘制波形
import wave
import matplotlib.pyplot as plt
import numpy as np

plt.rcParams[\'font.sans-serif\'] = [\'SimHei\']  # 用来正常显示中文标签
plt.rcParams[\'axes.unicode_minus\'] = False  # 用来正常显示符号

# 打开WAV音频
f = wave.open(r"C:\Windows\media\Windows Background.wav", "rb")

# 读取格式信息
# (声道数、量化位数、采样频率、采样点数、压缩类型、压缩类型的描述)
# (nchannels, sampwidth, framerate, nframes, comptype, compname)
params = f.getparams()
nchannels, sampwidth, framerate, nframes = params[:4]
# nchannels通道数 = 2
# sampwidth量化位数 = 2
# framerate采样频率 = 22050
# nframes采样点数 = 53395

# 读取nframes个数据，返回字符串格式
str_data = f.readframes(nframes)

f.close()

# 将字符串转换为数组，得到一维的short类型的数组
wave_data = np.fromstring(str_data, dtype=np.short)

# 赋值的归一化
wave_data = wave_data * 1.0 / (max(abs(wave_data)))

# 整合左声道和右声道的数据
wave_data = np.reshape(wave_data, [nframes, nchannels])
# wave_data.shape = (-1, 2)   # -1的意思就是没有指定,根据另一个维度的数量进行分割

# 最后通过采样点数和取样频率计算出每个取样的时间
time = np.arange(0, nframes) * (1.0 / framerate)

plt.figure()
# 左声道波形
plt.subplot(2, 1, 1)
plt.plot(time, wave_data[:, 0])
plt.xlabel("时间/s",fontsize=14)
plt.ylabel("幅度",fontsize=14)
plt.title("左声道",fontsize=14)
plt.grid()  # 标尺

plt.subplot(2, 1, 2)
# 右声道波形
plt.plot(time, wave_data[:, 1], c="g")
plt.xlabel("时间/s",fontsize=14)
plt.ylabel("幅度",fontsize=14)
plt.title("右声道",fontsize=14)

plt.tight_layout()  # 紧密布局
plt.show()

# Author:凌逆战
# -*- coding:utf-8 -*-
import wave
import numpy as np
import scipy.signal as signal

framerate = 44100   # 采样频率
time = 10           # 持续时间

t = np.arange(0, time, 1.0/framerate)

# 调用scipy.signal库中的chrip函数，
# 产生长度为10秒、取样频率为44.1kHz、100Hz到1kHz的频率扫描波
wave_data = signal.chirp(t, 100, time, 1000, method=\'linear\') * 10000

# 由于chrip函数返回的数组为float64型，
# 需要调用数组的astype方法将其转换为short型。
wave_data = wave_data.astype(np.short)

# 打开WAV音频用来写操作
f = wave.open(r"sweep.wav", "wb")

f.setnchannels(1)           # 配置声道数
f.setsampwidth(2)           # 配置量化位数
f.setframerate(framerate)   # 配置取样频率
comptype = "NONE"
compname = "not compressed"

# 也可以用setparams一次性配置所有参数
# outwave.setparams((1, 2, framerate, nframes,comptype, compname))

# 将wav_data转换为二进制数据写入文件
f.writeframes(wave_data.tostring())
f.close()

# Author:凌逆战
# -*- coding:utf-8 -*-
import wave
import numpy as np
import struct

f = wave.open(r"C:\Windows\media\Windows Background.wav", "rb")
params = f.getparams()
nchannels, sampwidth, framerate, nframes = params[:4]
strData = f.readframes(nframes)
waveData = np.fromstring(strData,dtype=np.int16)
f.close()
waveData = waveData*1.0/(max(abs(waveData)))

# wav文件写入
# 待写入wav的数据，这里仍然取waveData数据
outData = waveData
outwave = wave.open("write.wav", \'wb\')
nchannels = 1   # 通道数设置为1
sampwidth = 2   # 量化位数设置为2
framerate = 8000    # 采样频率8000
nframes = len(outData)    # 采样点数

comptype = "NONE"
compname = "not compressed"
outwave.setparams((nchannels, sampwidth, framerate, nframes,
    comptype, compname))

for i in outData:
        outwave.writeframes(struct.pack(\'h\', int(i * 64000 / 2)))

        # struct.pack(FMT, V1)将V1的值转换为FMT格式字符串
outwave.close()

import json
import numpy as np
from scipy.io.wavfile import write
import matplotlib.pyplot as plt

# 定义合成音调
def Synthetic_tone(freq, duration, amp=1.0, sampling_freq=44100):
    # 建立时间轴
    t = np.linspace(0, duration, duration * sampling_freq)
    # 构建音频信号
    audio = amp * np.sin(2 * np.pi * freq * t)
    return audio.astype(np.int16)


# json文件中包含一些音阶以及他们的频率
tone_map_file = \'tone_freq_map.json\'

# 读取频率映射文件
with open(tone_map_file, \'r\') as f:
    tone_freq_map = json.loads(f.read())
    print(tone_freq_map)
# {\'A\': 440, \'Asharp\': 466, \'B\': 494, \'C\': 523, \'Csharp\': 554, \'D\': 587, \'Dsharp\': 622, \'E\': 659, \'F\': 698, \'Fsharp\': 740, \'G\': 784, \'Gsharp\': 831}

# 设置生成G调的输入参数
input_tone = \'G\'
duration = 2             # seconds
amplitude = 10000        # 振幅
sampling_freq = 44100    # Hz
# 生成音阶
synthesized_tone = Synthetic_tone(tone_freq_map[input_tone], duration, amplitude, sampling_freq)

# 写入输出文件
write(\'output_tone.wav\', sampling_freq, synthesized_tone)

# 音阶及其连续时间
tone_seq = [(\'D\', 0.3), (\'G\', 0.6), (\'C\', 0.5), (\'A\', 0.3), (\'Asharp\', 0.7)]

# 构建基于和弦序列的音频信号
output = np.array([])
for item in tone_seq:
    input_tone = item[0]
    duration = item[1]
    synthesized_tone = Synthetic_tone(tone_freq_map[input_tone], duration, amplitude, sampling_freq)
    output = np.append(output, synthesized_tone, axis=0)

# 写入输出文件
write(\'output_tone_seq.wav\', sampling_freq, output)

# -*- coding: utf-8 -*-
import pyaudio
import wave

chunk = 1024

wf = wave.open(r"c:\WINDOWS\Media\Windows Background.wav", \'rb\')

p = pyaudio.PyAudio()

# 打开声音输出流
stream = p.open(format = p.get_format_from_width(wf.getsampwidth()),
                channels = wf.getnchannels(),
                rate = wf.getframerate(),
                output = True)

# 写声音输出流到声卡进行播放
while True:
    data = wf.readframes(chunk)
    if data == "":
        break
    stream.write(data)

stream.stop_stream()
stream.close()
p.terminate()   # 关闭PyAudio

\'\'\'
以SAMPLING_RATE为采样频率，
每次读入一块有NUM_SAMPLES个采样点的数据块，
当读入的采样数据中有COUNT_NUM个值大于LEVEL的取样的时候，
将采样数据保存进WAV文件，
一旦开始保存数据，所保存的数据长度最短为SAVE_LENGTH个数据块。

从声卡读入的数据和从WAV文件读入的类似，都是二进制数据，
由于我们用paInt16格式(16bit的short类型)保存采样值，
因此将它自己转换为dtype为np.short的数组。
\'\'\'


from pyaudio import PyAudio, paInt16
import numpy as np
import wave

# 将data中的数据保存到名为filename的WAV文件中
def save_wave_file(filename, data):
    wf = wave.open(filename, \'wb\')
    wf.setnchannels(1)          # 单通道
    wf.setsampwidth(2)          # 量化位数
    wf.setframerate(SAMPLING_RATE)  # 设置采样频率
    wf.writeframes(b"".join(data))  # 写入语音帧
    wf.close()


NUM_SAMPLES = 2000      # pyAudio内部缓存块的大小
SAMPLING_RATE = 8000    # 取样频率
LEVEL = 1500           # 声音保存的阈值，小于这个阈值不录
COUNT_NUM = 20 # 缓存快类如果有20个大于阈值的取样则记录声音
SAVE_LENGTH = 8 # 声音记录的最小长度：SAVE_LENGTH * NUM_SAMPLES 个取样

# 开启声音输入
pa = PyAudio()
stream = pa.open(format=paInt16, channels=1, rate=SAMPLING_RATE, input=True,
                frames_per_buffer=NUM_SAMPLES)

save_count = 0  # 用来计数
save_buffer = []    #

while True:
    # 读入NUM_SAMPLES个取样
    string_audio_data = stream.read(NUM_SAMPLES)
    # 将读入的数据转换为数组
    audio_data = np.fromstring(string_audio_data, dtype=np.short)
    # 计算大于LEVEL的取样的个数
    large_sample_count = np.sum( audio_data > LEVEL )
    print(np.max(audio_data))
    # 如果个数大于COUNT_NUM，则至少保存SAVE_LENGTH个块
    if large_sample_count > COUNT_NUM:
        save_count = SAVE_LENGTH
    else:
        save_count -= 1

    if save_count < 0:
        save_count = 0

    if save_count > 0:
        # 将要保存的数据存放到save_buffer中
        save_buffer.append( string_audio_data )
    else:
        # 将save_buffer中的数据写入WAV文件，WAV文件的文件名是保存的时刻
        if len(save_buffer) > 0:
            filename = "recorde" + ".wav"
            save_wave_file(filename, save_buffer)
            print(filename, "saved")
            break

import matplotlib.pyplot as plt
import scipy.signal as signal
plt.figure(figsize=(6,2))
plt.plot(signal.hanning(512))
plt.show()

import numpy as np
import wave
import os
#import math
 
def enframe(signal, nw, inc):
    \'\'\'将音频信号转化为帧。
    参数含义：
    signal:原始音频型号
    nw:每一帧的长度(这里指采样点的长度，即采样频率乘以时间间隔)
    inc:相邻帧的间隔（同上定义）
    \'\'\'
    signal_length=len(signal) #信号总长度
    if signal_length<=nw: #若信号长度小于一个帧的长度，则帧数定义为1
        nf=1
    else: #否则，计算帧的总长度
        nf=int(np.ceil((1.0*signal_length-nw+inc)/inc))
    pad_length=int((nf-1)*inc+nw) #所有帧加起来总的铺平后的长度
    zeros=np.zeros((pad_length-signal_length,)) #不够的长度使用0填补，类似于FFT中的扩充数组操作
    pad_signal=np.concatenate((signal,zeros)) #填补后的信号记为pad_signal
    indices=np.tile(np.arange(0,nw),(nf,1))+np.tile(np.arange(0,nf*inc,inc),(nw,1)).T  #相当于对所有帧的时间点进行抽取，得到nf*nw长度的矩阵
    indices=np.array(indices,dtype=np.int32) #将indices转化为矩阵
    frames=pad_signal[indices] #得到帧信号
#    win=np.tile(winfunc(nw),(nf,1))  #window窗函数，这里默认取1
#    return frames*win   #返回帧信号矩阵
    return frames
def wavread(filename):
    f = wave.open(filename,\'rb\')
    params = f.getparams()
    nchannels, sampwidth, framerate, nframes = params[:4]
    strData = f.readframes(nframes)#读取音频，字符串格式
    waveData = np.fromstring(strData,dtype=np.int16)#将字符串转化为int
    f.close()
    waveData = waveData*1.0/(max(abs(waveData)))#wave幅值归一化
    waveData = np.reshape(waveData,[nframes,nchannels]).T
    return waveData
 
filepath = "./data/" #添加路径
dirname= os.listdir(filepath) #得到文件夹下的所有文件名称 
filename = filepath+dirname[0]
data = wavread(filename)
nw = 512
inc = 128
Frame = enframe(data[0], nw, inc) 

没有加窗的语音分帧

def enframe(signal, nw, inc, winfunc):
    \'\'\'将音频信号转化为帧。
    参数含义：
    signal:原始音频型号
    nw:每一帧的长度(这里指采样点的长度，即采样频率乘以时间间隔)
    inc:相邻帧的间隔（同上定义）
    \'\'\'
    signal_length=len(signal) #信号总长度
    if signal_length<=nw: #若信号长度小于一个帧的长度，则帧数定义为1
        nf=1
    else: #否则，计算帧的总长度
        nf=int(np.ceil((1.0*signal_length-nw+inc)/inc))
    pad_length=int((nf-1)*inc+nw) #所有帧加起来总的铺平后的长度
    zeros=np.zeros((pad_length-signal_length,)) #不够的长度使用0填补，类似于FFT中的扩充数组操作
    pad_signal=np.concatenate((signal,zeros)) #填补后的信号记为pad_signal
    indices=np.tile(np.arange(0,nw),(nf,1))+np.tile(np.arange(0,nf*inc,inc),(nw,1)).T  #相当于对所有帧的时间点进行抽取，得到nf*nw长度的矩阵
    indices=np.array(indices,dtype=np.int32) #将indices转化为矩阵
    frames=pad_signal[indices] #得到帧信号
    win=np.tile(winfunc,(nf,1))  #window窗函数，这里默认取1
    return frames*win   #返回帧信号矩阵

import numpy as np
from scipy.io import wavfile
import matplotlib.pyplot as plt

sampling_freq, audio = wavfile.read(r"C:\Windows\media\Windows Background.wav")   # 读取文件

audio = audio / np.max(audio)   # 归一化，标准化

# 应用傅里叶变换
fft_signal = np.fft.fft(audio)
print(fft_signal)
# [-0.04022912+0.j         -0.04068997-0.00052721j -0.03933007-0.00448355j
#  ... -0.03947908+0.00298096j -0.03933007+0.00448355j -0.04068997+0.00052721j]

fft_signal = abs(fft_signal)
print(fft_signal)
# [0.04022912 0.04069339 0.0395848  ... 0.08001755 0.09203427 0.12889393]

# 建立时间轴
Freq = np.arange(0, len(fft_signal))

# 绘制语音信号的
plt.figure()
plt.plot(Freq, fft_signal, color=\'blue\')
plt.xlabel(\'Freq (in kHz)\')
plt.ylabel(\'Amplitude\')
plt.show()

import numpy as np
import matplotlib.pyplot as plt
from scipy.io import wavfile
from python_speech_features import mfcc, logfbank

# 读取输入音频文件
sampling_freq, audio = wavfile.read("input_freq.wav")

# 提取MFCC和滤波器组特征
mfcc_features = mfcc(audio, sampling_freq)
filterbank_features = logfbank(audio, sampling_freq)

print(\'\nMFCC:\n窗口数 =\', mfcc_features.shape[0])
print(\'每个特征的长度 =\', mfcc_features.shape[1])
print(\'\nFilter bank:\n窗口数 =\', filterbank_features.shape[0])
print(\'每个特征的长度 =\', filterbank_features.shape[1])

# 画出特征图，将MFCC可视化。转置矩阵，使得时域是水平的
mfcc_features = mfcc_features.T
plt.matshow(mfcc_features)
plt.title(\'MFCC\')
# 将滤波器组特征可视化。转置矩阵，使得时域是水平的
filterbank_features = filterbank_features.T
plt.matshow(filterbank_features)
plt.title(\'Filter bank\')

plt.show()

import wave
import matplotlib.pyplot as plt
import numpy as np

f = wave.open(r"C:\Windows\media\Windows Background.wav", "rb")
params = f.getparams()
nchannels, sampwidth, framerate, nframes = params[:4]
strData = f.readframes(nframes)#读取音频，字符串格式
waveData = np.fromstring(strData,dtype=np.int16)#将字符串转化为int
waveData = waveData*1.0/(max(abs(waveData)))#wave幅值归一化
waveData = np.reshape(waveData,[nframes,nchannels]).T
f.close()

plt.specgram(waveData[0],Fs = framerate, scale_by_freq = True, sides = \'default\')
plt.ylabel(\'Frequency(Hz)\')
plt.xlabel(\'Time(s)\')
plt.show()

[Y,FS]=audioread(\'p225_355_wb.wav\');

% specgram(Y,2048,44100,2048,1536);
    %Y1为波形数据
    %FFT帧长2048点(在44100Hz频率时约为46ms)
    %采样频率44.1KHz
    %加窗长度，一般与帧长相等
    %帧重叠长度，此处取为帧长的3/4
specgram(Y,2048,FS,2048,1536);
xlabel(\'时间(s)\')
ylabel(\'频率(Hz)\')
title(\'语谱图\')

import os
import numpy as np
import scipy.io.wavfile as wf
import python_speech_features as sf
import hmmlearn.hmm as hl


# 1. 读取training文件夹中的训练音频样本，每个音频对应一个mfcc矩阵，每个mfcc都有一个类别(apple...)
def search_file(directory):
    """
    :param directory: 训练音频的路径
    :return: 字典{\'apple\':[url, url, url ... ], \'banana\':[...]}
    """
    # 使传过来的directory匹配当前操作系统
    directory = os.path.normpath(directory)
    objects = {}
    # curdir：当前目录
    # subdirs: 当前目录下的所有子目录
    # files: 当前目录下的所有文件名
    for curdir, subdirs, files in os.walk(directory):
        for file in files:
            if file.endswith(\'.wav\'):
                label = curdir.split(os.path.sep)[-1]  # os.path.sep为路径分隔符
                if label not in objects:
                    objects[label] = []
                # 把路径添加到label对应的列表中
                path = os.path.join(curdir, file)
                objects[label].append(path)
    return objects


# 读取训练集数据
train_samples = search_file(\'../machine_learning_date/speeches/training\')

"""
2. 把所有类别为apple的mfcc合并在一起，形成训练集。
    训练集:
    train_x：[mfcc1,mfcc2,mfcc3,...],[mfcc1,mfcc2,mfcc3,...]...
    train_y：[apple],[banana]...
由上述训练集样本可以训练一个用于匹配apple的HMM。"""

train_x, train_y = [], []
# 遍历字典
for label, filenames in train_samples.items():
    # [(\'apple\', [\'url1,,url2...\'])
    # [("banana"),("url1,url2,url3...")]...
    mfccs = np.array([])
    for filename in filenames:
        sample_rate, sigs = wf.read(filename)
        mfcc = sf.mfcc(sigs, sample_rate)
        if len(mfccs) == 0:
            mfccs = mfcc
        else:
            mfccs = np.append(mfccs, mfcc, axis=0)
    train_x.append(mfccs)
    train_y.append(label)

# 3.训练模型，有7个句子，创建了7个模型
models = {}
for mfccs, label in zip(train_x, train_y):
    model = hl.GaussianHMM(n_components=4, covariance_type=\'diag\', n_iter=1000)
    models[label] = model.fit(mfccs)  # # {\'apple\':object, \'banana\':object ...}

"""
4. 读取testing文件夹中的测试样本，
    测试集数据：
        test_x  [mfcc1, mfcc2, mfcc3...]
        test_y  [apple, banana, lime]
"""
test_samples = search_file(\'../machine_learning_date/speeches/testing\')

test_x, test_y = [], []
for label, filenames in test_samples.items():
    mfccs = np.array([])
    for filename in filenames:
        sample_rate, sigs = wf.read(filename)
        mfcc = sf.mfcc(sigs, sample_rate)
        if len(mfccs) == 0:
            mfccs = mfcc
        else:
            mfccs = np.append(mfccs, mfcc, axis=0)
    test_x.append(mfccs)
    test_y.append(label)


# 5.测试模型
#    1. 分别使用7个HMM模型，对测试样本计算score得分。
#    2. 取7个模型中得分最高的模型所属类别作为预测类别。
pred_test_y = []
for mfccs in test_x:
    # 判断mfccs与哪一个HMM模型更加匹配
    best_score, best_label = None, None
    # 遍历7个模型
    for label, model in models.items():
        score = model.score(mfccs)
        if (best_score is None) or (best_score < score):
            best_score = score
            best_label = label
    pred_test_y.append(best_label)

print(test_y)   # [\'apple\', \'banana\', \'kiwi\', \'lime\', \'orange\', \'peach\', \'pineapple\']
print(pred_test_y)  # [\'apple\', \'banana\', \'kiwi\', \'lime\', \'orange\', \'peach\', \'pineapple\']