python语音识别基于hmmlearn库（隐马尔科夫模型）

import os
import numpy as np
from scipy.io import wavfile # 仅限于wav音频文件
from hmmlearn import hmm
from python_speech_features import mfcc

创建HMM类详见：https://www.cnblogs.com/pinard/p/7001397.html

MFCC详见：https://blog.csdn.net/xmdxcsj/article/details/51228791

class HMMTrainer(object):
def init(self, model_name=’GaussianHMM’, n_components=4, cov_type=’diag’, n_iter=1000):
self.model_name = model_name #模型名称 hmmlearn实现了三种HMM模型类，GaussianHMM和GMMHMM是连续观测状态的HMM模型，MultinomialHMM是离散观测状态的模型
self.n_components = n_components #隐藏状态个数
self.cov_type = cov_type #转移矩阵协方差类型
self.n_iter = n_iter #训练迭代次数
self.models = []

    if self.model_name == 'GaussianHMM':
        self.model = hmm.GaussianHMM(n_components=self.n_components, 
                covariance_type=self.cov_type, n_iter=self.n_iter)
    else:
        raise TypeError('Invalid model type')

# X是2维numpy数组每行13列
def train(self, X):
    np.seterr(all='ignore')
    self.models.append(self.model.fit(X))

# 测试输入的模型得分
def get_score(self, input_data):
    return self.model.score(input_data)

if name==’main‘:
# 输入文件夹路径（输入路径包含子文件夹，每个子文件夹名为音频文件名称，内含多个同音频，供模型训练）
input_folder = args.input_folder

hmm_models = []

# 遍历输入路径
for dirname in os.listdir(input_folder):
    # 获取子文件名 
    subfolder = os.path.join(input_folder, dirname)
    # 跳过非文件夹
    if not os.path.isdir(subfolder): 
        continue

    # 子文件夹名作为标签
    label = subfolder[subfolder.rfind('/') + 1:]

    # 初始化变量 
    X = np.array([])
    y_words = []

    # 迭代音频文件 (每类文件即：子文件夹文件，留下一个做测试用)
    for filename in [x for x in os.listdir(subfolder) if x.endswith('.wav')][:-1]:
        # 读取音频文件
        filepath = os.path.join(subfolder, filename)
        sampling_freq, audio = wavfile.read(filepath)

        # 解析MFCC特征（MFCC：梅尔频率倒谱系数）
        mfcc_features = mfcc(audio, sampling_freq)

        # 添加mfcc_features
        if len(X) == 0:
            X = mfcc_features
        else:
            X = np.append(X, mfcc_features, axis=0)

        # 添加标签
        y_words.append(label)

    print 'X.shape =', X.shape
    # 训练并保存模型
    # 创建HMM类 
    hmm_trainer = HMMTrainer()
    # 训练模型
    hmm_trainer.train(X)
    hmm_models.append((hmm_trainer, label))
    hmm_trainer = None

# 测试文件路径
input_files = [
        'data/pineapple/pineapple15.wav',
        'data/orange/orange15.wav',
        'data/apple/apple15.wav',
        'data/kiwi/kiwi15.wav'
        ]

# 读取测试文件
for input_file in input_files:
    # 读取测试文件
    # sampling_freq采样频率
    sampling_freq, audio = wavfile.read(input_file)

    # 提取MFCC特征
    mfcc_features = mfcc(audio, sampling_freq)

    # 定义模型得分，输出标签
    max_score = None
    output_label = None

    # 迭代所有模型 
    # 得分最高的模型对应的标签，即为输出标签（识别值）
    for item in hmm_models:
        hmm_model, label = item
        score = hmm_model.get_score(mfcc_features)
        if score > max_score:
            max_score = score
            output_label = label

    # 打印输出
    print "\nTrue:", input_file[input_file.find('/')+1:input_file.rfind('/')]
    print "Predicted:", output_label 

转载请注明出处