超简单的婴儿哭声检测实现方案--python版

1. 构建项目

项目结构

└─audio_data ## 音频文件
├─mp3
├─test
└─wav ## 训练音频源文件 每个目录代表不同的标签，自己定义和随意增加，cry 目录存放的是婴儿的哭声
├─cry
├─non_cry
└─other
└─get-model.py
└─main.py
└─requirements.txt

get-model.py

## get-model.py
import os
import librosa
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score, classification_report
from joblib import dump, load # 使用 joblib 直接导入 dump 和 load
from packaging import version # 用于版本比较

# 加载音频文件
def load_audio_file(file_path):
signal, sample_rate = librosa.load(file_path, sr=None, mono=True)
return signal, sample_rate

# 提取 MFCC 特征
def extract_features(file_path, sample_rate):
signal, _ = load_audio_file(file_path)
mfccs = librosa.feature.mfcc(y=signal, sr=sample_rate, n_mfcc=13)
mfccs_scaled_features = np.mean(mfccs.T, axis=0)
return mfccs_scaled_features

# 数据目录
data_dir = './audio_data/wav' # 不使用反斜杠
# 假设类别包括 'cry', 'non_cry', 'other' 等
labels = {'cry': 1, 'non_cry': 0, 'other': 2} # 添加其他类别

# 初始化特征和标签列表
X, y = [], []

# 遍历数据目录中的文件
for root, dirs, files in os.walk(data_dir):
print(f"Processing directory: {root}")
for file in files:
print(f"File found: {os.path.join(root, file)}")
if file.endswith('.wav'):
file_label = os.path.basename(root).split('_')[0] # 获取文件夹名称
print(f"File label: {file_label}")
if file_label in labels:
file_path = os.path.join(root, file)
print(f"Processing file: {file_path}")
try:
feature = extract_features(file_path, sample_rate=22050)
X.append(feature)
y.append(labels[file_label])
print(f"Feature extracted successfully.")
print(f"Feature: {feature}")
print(f"Label: {labels[file_label]}")
except Exception as e:
print(f"Error processing file {file_path}: {e}")

# 检查是否有数据
if len(X) == 0 or len(y) == 0:
raise ValueError("No valid data found in the directory.")

# 转换为 NumPy 数组
X = np.array(X)
y = np.array(y)

# 打印数据形状以确认是否正确加载
print(f"Shape of X: {X.shape}")
print(f"Shape of y: {y.shape}")
print(f"X: {X}")
print(f"y: {y}")

# 检查标签分布
unique_labels, counts = np.unique(y, return_counts=True)
print(f"Unique labels: {unique_labels}")
print(f"Counts: {counts}")

# 如果标签数量少于2，则抛出异常
if len(unique_labels) < 2:
raise ValueError("The dataset must contain at least two different labels.")

# 划分数据集
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=42)

# 标准化特征
scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)

# 训练模型
# 使用 'ovr'（一对多）或 'ovo'（一对一）策略来支持多类分类
model = SVC(kernel='linear', C=1, decision_function_shape='ovr') # 支持多类分类
model.fit(X_train, y_train)

# 评估模型
y_pred = model.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
report = classification_report(y_test, y_pred)

print(f"Accuracy: {accuracy}")
print(report)

# 获取 sklearn 的版本号
from sklearn import __version__ as sklearn_version

# 检查并使用正确的 joblib 方法
if version.parse(sklearn_version) >= version.parse("0.24"):
dump(model, 'model.pkl')
dump(scaler, 'scaler.pkl')
else:
joblib.dump(model, 'model.pkl')
joblib.dump(scaler, 'scaler.pkl')

print("Model and scaler saved.")

main.py

## main.py
import os
import librosa
import numpy as np
from sklearn.preprocessing import StandardScaler
from joblib import load # 使用 joblib 直接导入 load

# 加载音频文件
def load_audio_file(file_path):
signal, sample_rate = librosa.load(file_path, sr=None, mono=True)
return signal, sample_rate

# 提取 MFCC 特征
def extract_features(file_path, sample_rate):
signal, _ = load_audio_file(file_path)
mfccs = librosa.feature.mfcc(y=signal, sr=sample_rate, n_mfcc=13)
mfccs_scaled_features = np.mean(mfccs.T, axis=0)
return mfccs_scaled_features

# 加载模型和标量
# 假设你有一个名为 'model.pkl' 的训练好的模型文件
model = load('model.pkl') # 加载模型
scaler = load('scaler.pkl') # 加载标量

# 指定测试文件目录
test_dir = './audio_data/test'

# 遍历指定目录下的所有 .mp3 文件
for root, dirs, files in os.walk(test_dir):
for file in files:
if file.endswith('.mp3') or file.endswith('.wav'):
file_path = os.path.join(root, file)
print(f"Processing file: {file_path}")

try:
# 提取特征
features = extract_features(file_path, sample_rate=22050)

# 标准化特征
features_scaled = scaler.transform([features])

# 进行预测
prediction = model.predict(features_scaled)

# 输出预测结果
print(f"文件名: {file} --- 结果值: {prediction[0]}")
except Exception as e:
print(f"Error processing file {file_path}: {e}")

librosa
numpy
scipy
sklearn
joblib

2. 安装依赖

安装环境

1. cmd或者powershell 直接执行以下命令，Windows10 会自动跳转到应用市场，根据提示安装即可

Python

2. 安装依赖包

pip3 install -r requirements.txt

3.训练模型

python3 get-model.py ## 训练完成会生成两个文件： model.pkl scaler.pkl

4.测试模型

python3 main.py