{"id":486,"date":"2023-05-08T10:52:28","date_gmt":"2023-05-08T02:52:28","guid":{"rendered":"https:\/\/zhaocunwei.co.uk\/?p=486"},"modified":"2023-05-08T10:54:11","modified_gmt":"2023-05-08T02:54:11","slug":"the-project-uses-voice-interaction","status":"publish","type":"post","link":"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/","title":{"rendered":"\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09"},"content":{"rendered":"<h1>\u524d\u8a00<\/h1>\n<p>\u6700\u8fd1\u9879\u76ee\u4e2d\u8981\u4f7f\u7528\u8bed\u97f3\u8bc6\u522b\uff0c\u67e5\u9605\u8d44\u6599\u627e\u5230\u4e86\u4e00\u4e2a\u795e\u5947\u7684\u6a21\u578b\uff1aCRNN \uff0c\u901a\u8fc7\u7f16\u5199\u6587\u7ae0\uff0c\u53ef\u4ee5\u6559\u60a8\u5927\u6982\u77e5\u9053\u8fd9\u4e2aCRNN\u6a21\u578b\u7684\u5177\u4f53\u7528\u6cd5\uff0c\ud83d\ude00<\/p>\n<h1>O(\u2229_\u2229)O\u54c8\u54c8~<\/h1>\n<p>CRNN\u662f\u4ec0\u4e48?<\/p>\n<p>CRNN\u662fConvolutional Neural Network(\u5377\u79ef\u795e\u7ecf\u7f51\u7edc)\u548cRecurrent Neural Network(\u9012\u5f52\u795e\u7ecf\u7f51\u7edc)\u7684\u7ed3\u5408,\u662f\u4e00\u4e2a\u5e8f\u5217\u5230\u5e8f\u5217(Sequence to Sequence)\u7684\u6a21\u578b\u3002<\/p>\n<p>\u901a\u8fc7\u8d44\u6599\u67e5\u8be2\u4e86\u89e3\u5230\uff1a\u5b83\u901a\u8fc7CNN\u63d0\u53d6\u5e8f\u5217\u6570\u636e\u7684\u7279\u5f81,\u518d\u7528RNN(\u901a\u5e38\u662fLSTM\u6216GRU)\u6765\u5efa\u6a21\u957f\u4f9d\u8d56\u5173\u7cfb,\u5b9e\u73b0\u5e8f\u5217\u6570\u636e\u7684\u5206\u7c7b\u6216\u9884\u6d4b\u3002<\/p>\n<p>\u672c\u7bc7\u6587\u7ae0\u4e3b\u8981\u901a\u8fc7\u4e00\u4e0b\u51e0\u4e2a\u65b9\u9762\uff0c CRNN\u7684\u7ed3\u6784\u4e00\u4e2a\u5178\u578b\u7684CRNN\u6a21\u578b\u5305\u542b\u4ee5\u4e0b\u5c42:<\/p>\n<p>\u5377\u79ef\u5c42(Convolution Layer):\u901a\u8fc7\u591a\u5c42\u5377\u79ef\u63d0\u53d6\u4e0d\u540c\u7c92\u5ea6\u7684\u7279\u5f81,\u8d77\u5230\u7279\u5f81\u63d0\u53d6\u7684\u4f5c\u7528\u3002<\/p>\n<p>\u6c60\u5316\u5c42(Pooling Layer):\u7528\u4e8e\u964d\u7ef4,\u51cf\u5c11\u8ba1\u7b97\u590d\u6742\u5ea6\u3002<br \/>\n\u9012\u5f52\u5c42(Recurrent Layer):\u901a\u5e38\u662fLSTM\u6216GRU,\u7528\u4e8e\u5efa\u6a21\u957f\u5e8f\u5217\u4f9d\u8d56\u5173\u7cfb\u3002<br \/>\n\u5168\u8fde\u63a5\u5c42(Fully Connected Layer):\u7528\u4e8e\u6700\u7ec8\u7684\u5206\u7c7b\u6216\u56de\u5f52\u9884\u6d4b\u3002<\/p>\n<p>\u5728\u5b9e\u9645\u9879\u76ee\u5f53\u4e2d\u6211\u4eec\uff0c\u5982\u4f55\u4f7f\u7528CRNN\u6a21\u578b\uff0c\ud83d\udc47\u7ed9\u5927\u5bb6\u5206\u4eab\u4e00\u4e0b\uff0c\u4f7f\u7528CRNN\u6a21\u578b\u901a\u5e38\u5206\u4e3a\u4ee5\u4e0b\u51e0\u6b65:<\/p>\n<p>\u200b               <font color=red>\u5b9a\u4e49CRNN\u6a21\u578b\u7ed3\u6784:\u6dfb\u52a0\u5377\u79ef\u5c42\u3001\u6c60\u5316\u5c42\u3001LSTM\u5c42\u548c\u5168\u8fde\u63a5\u5c42\u3002<\/font><\/p>\n<p>\u200b               <\/p>\n<pre><code>import tensorflow as tf\nfrom tensorflow.keras.layers import Conv2D, MaxPool2D, LSTM, Dense\n\nclass CRNNModel(tf.keras.Model):\n    def __init__(self):\n        super().__init__()\n        # \u5377\u79ef\u5c42\n        self.conv1 = Conv2D(32, (3, 3), activation=&#039;relu&#039;)\n        self.conv2 = Conv2D(64, (3, 3), activation=&#039;relu&#039;)\n        # \u6c60\u5316\u5c42\n        self.pool1 = MaxPool2D(pool_size=(2, 2))\n        self.pool2 = MaxPool2D(pool_size=(2, 2))\n        # LSTM\u5c42\n        self.lstm = LSTM(128) \n        # \u5168\u8fde\u63a5\u5c42\n        self.fc = Dense(10, activation=&#039;softmax&#039;)\n\n    def call(self, inputs):\n        # \u5377\u79ef\n        x = self.conv1(inputs)\n        x = self.pool1(x)\n        x = self.conv2(x)\n        x = self.pool2(x)\n\n        # LSTM\n        x = tf.reshape(x, [-1, x.shape[1], x.shape[2]*x.shape[3]])\n        x = self.lstm(x)\n\n        # \u5168\u8fde\u63a5\n        output = self.fc(x)\n\n        return output<\/code><\/pre>\n<p>\u8fd9\u4e2a\u6a21\u578b\u5305\u542b:<\/p>\n<p>2\u4e2aConv2D\u5c42:\u7b2c\u4e00\u4e2a\u5377\u79ef\u6838\u4e3a32,\u7b2c\u4e8c\u4e2a\u4e3a64\u3002\u7528\u6765\u63d0\u53d6\u7279\u5f81\u3002<br \/>\n2\u4e2aMaxPool2D\u5c42:\u7528\u4e8e\u6c60\u5316,\u51cf\u5c11\u8ba1\u7b97\u91cf\u3002<br \/>\n1\u4e2aLSTM\u5c42:\u7528\u4e8e\u5efa\u6a21\u957f\u5e8f\u5217\u4f9d\u8d56,lstm_size\u4e3a128\u3002<br \/>\n1\u4e2aDense\u5c42:\u4f5c\u4e3a\u8f93\u51fa\u5c42,\u795e\u7ecf\u5143\u6570\u91cf\u4e3a10,\u4f7f\u7528softmax\u6fc0\u6d3b\u8fdb\u884c\u591a\u5206\u7c7b\u3002\u5728call\u65b9\u6cd5\u4e2d,\u5148\u901a\u8fc7\u4e24\u4e2a\u5377\u79ef\u5c42\u548c\u6c60\u5316\u5c42\u63d0\u53d6\u7279\u5f81,\u7136\u540e\u5c06\u7279\u5f81\u5e8f\u5217\u8f93\u5165\u5230LSTM\u5c42,\u6700\u540e\u4f7f\u7528Dense\u5c42\u4f5c\u4e3a\u8f93\u51fa,\u5f97\u5230\u5206\u7c7b\u9884\u6d4b\u7ed3\u679c\u3002<\/p>\n<p>\u200b               <font color=red>\u7f16\u8bd1\u6a21\u578b:\u9009\u62e9\u635f\u5931\u51fd\u6570\u548c\u4f18\u5316\u5668,\u5b9a\u4e49\u8bc4\u6d4b\u6307\u6807\u3002<\/font><\/p>\n<p>\u200b                           \u5f53\u6211\u4eec\u5728\u6784\u5efa\u597dCRNN\u6a21\u578b\u7ed3\u6784\u540e,\u4e0b\u4e00\u6b65\u662f\u7f16\u8bd1\u6a21\u578b\u3002<\/p>\n<pre><code>                    \u7f16\u8bd1\u6a21\u578b\u4e3b\u8981\u662f:<\/code><\/pre>\n<ol>\n<li>\u9009\u62e9\u635f\u5931\u51fd\u6570(loss function):\u8861\u91cf\u6a21\u578b\u7684\u9884\u6d4b\u7ed3\u679c\u548c\u771f\u5b9e\u6807\u7b7e\u7684\u5dee\u5f02,\u5e38\u7528\u7684\u6709\u4ea4\u53c9\u71b5\u635f\u5931\u51fd\u6570cross entropy loss\u3001\u5747\u65b9\u8bef\u5deeloss\u7b49\u3002<\/li>\n<\/ol>\n<p>\u200b        \ud83d\udc47\u662f\u65e5\u5e38\u9879\u76ee\u5f53\u4e2d\u5e94\u7528\u5230\u7684\u51fd\u6570\uff1a<\/p>\n<p>\u200b             <\/p>\n<ol>\n<li>\u4ea4\u53c9\u71b5\u635f\u5931\u51fd\u6570(Cross Entropy Loss):<\/li>\n<\/ol>\n<pre><code>python\n# PyTorch\ncriterion = nn.CrossEntropyLoss()\n\n# TensorFlow\nloss = tf.keras.losses.CategoricalCrossentropy()<\/code><\/pre>\n<p>\u4ea4\u53c9\u71b5\u9002\u7528\u4e8e\u591a\u5206\u7c7b\u4efb\u52a1,\u5b83\u4f1a\u7ed9\u9519\u5206\u6837\u672c\u4ee5\u6bd4\u8f83\u5927\u7684\u635f\u5931\u60e9\u7f5a\u3002<\/p>\n<ol start=\"2\">\n<li>\u5747\u65b9\u8bef\u5dee\u635f\u5931\u51fd\u6570(Mean Squared Error Loss):<\/li>\n<\/ol>\n<pre><code>python \n# PyTorch\ncriterion = nn.MSELoss()\n\n# TensorFlow\nloss = tf.keras.losses.MeanSquaredError()<\/code><\/pre>\n<p>MSE\u9002\u7528\u4e8e\u56de\u5f52\u4efb\u52a1,\u5b83\u4f7f\u7528\u6837\u672c\u9884\u6d4b\u503c\u548c\u771f\u5b9e\u503c\u4e4b\u5dee\u7684\u5e73\u65b9\u4f5c\u4e3a\u635f\u5931\u3002<\/p>\n<ol start=\"3\">\n<li>\u4e8c\u5206\u7c7b\u7684\u4ea4\u53c9\u71b5\u635f\u5931\u51fd\u6570(Binary Cross Entropy Loss):<\/li>\n<\/ol>\n<pre><code>python\n# PyTorch\ncriterion = nn.BCELoss()\n\n# TensorFlow \nloss = tf.keras.losses.BinaryCrossentropy()<\/code><\/pre>\n<p>Binary Cross Entropy\u7528\u4e8e\u4e8c\u5206\u7c7b\u4efb\u52a1,\u5b83\u4f7f\u7528\u6982\u7387\u7684\u5bf9\u6570\u6765\u8ba1\u7b97\u4ea4\u53c9\u71b5\u3002<\/p>\n<ol start=\"4\">\n<li>\u81ea\u5b9a\u4e49\u635f\u5931\u51fd\u6570:<\/li>\n<\/ol>\n<pre><code>python\n# PyTorch\ndef my_loss_fn(outputs, targets):\n    ...\n    return loss\n\ncriterion = my_loss_fn\n\n# TensorFlow\ndef my_loss_fn(y_true, y_pred):\n    ...\n    return loss\n\nloss = tf.keras.losses.LossFunctionWrapper(my_loss_fn)<\/code><\/pre>\n<p>\u6211\u4eec\u4e5f\u53ef\u4ee5\u6839\u636e\u81ea\u5df1\u5b9e\u9645\u9879\u76ee\u4efb\u52a1\u7684\u9700\u6c42\u5b9a\u4e49\u81ea\u5df1\u7684\u635f\u5931\u51fd\u6570\u3002O(\u2229_\u2229)O\u54c8\u54c8~<\/p>\n<ol start=\"2\">\n<li>\n<p>\u9009\u62e9\u4f18\u5316\u5668(optimizer):\u66f4\u65b0\u6a21\u578b\u53c2\u6570\u4ee5\u51cf\u5c0f\u635f\u5931\u51fd\u6570,\u5e38\u7528\u7684\u6709Adam\u3001SGD\u7b49\u3002<\/p>\n<p><font color=red>\u5212\u91cd\u70b9<\/font><\/p>\n<p>\ud83d\udc6c\ud83c\udffb\u6211\u4eec\u77e5\u9053\u4f18\u5316\u5668\u7684\u4e3b\u8981\u4f5c\u7528\u662f\u901a\u8fc7\u8fed\u4ee3\u66f4\u65b0\u6a21\u578b\u53c2\u6570\u4f7f\u635f\u5931\u51fd\u6570\u6700\u5c0f\u5316\u5bf9\u5427\uff0c\ud83d\ude00\u73b0\u5728\u5206\u4eab\u4e00\u4e0b\u5e38\u7528\u7684\u4f18\u5316\u5668\u6709:<\/p>\n<ol>\n<li>Adam\u4f18\u5316\u5668:<\/li>\n<\/ol>\n<pre><code>python\n# PyTorch\noptimizer = torch.optim.Adam(model.parameters(), lr=0.001)\n\n# TensorFlow \noptimizer = tf.keras.optimizers.Adam(learning_rate=0.001)<\/code><\/pre>\n<p>Adam\u4f18\u5316\u5668\u6bd4\u8f83\u5e38\u7528,\u5b83\u7ed3\u5408\u4e86Momentum\u548cRMSProp,\u5bf9\u4e0d\u540c\u53c2\u6570\u4f7f\u7528\u4e0d\u540c\u7684\u5b66\u4e60\u7387,\u6bd4\u8f83\u9002\u7528\u4e8e\u9ad8\u5ea6\u975e\u7ebf\u6027\u7684\u95ee\u9898\u3002<\/p>\n<ol start=\"2\">\n<li>SGD\u4f18\u5316\u5668:<\/li>\n<\/ol>\n<pre><code>python\n# PyTorch\noptimizer = torch.optim.SGD(model.parameters(), lr=0.001, momentum=0.9)\n\n# TensorFlow\noptimizer = tf.keras.optimizers.SGD(learning_rate=0.001, momentum=0.9)<\/code><\/pre>\n<p>SGD\u4ee3\u8868Stochastic Gradient Descent,\u5b83\u4ee5\u4e00\u5b9a\u7684\u5b66\u4e60\u7387\u66f4\u65b0\u53c2\u6570,\u6bd4\u8f83\u7b80\u5355\u3002momentum\u7528\u4e8e\u589e\u52a0\u7a33\u5b9a\u6027\u548c\u6536\u655b\u901f\u5ea6\u3002<\/p>\n<ol start=\"3\">\n<li>RMSprop\u4f18\u5316\u5668:<\/li>\n<\/ol>\n<pre><code>python\n# PyTorch \noptimizer = torch.optim.RMSprop(model.parameters(), lr=0.001, alpha=0.99)\n\n# TensorFlow\noptimizer = tf.keras.optimizers.RMSprop(learning_rate=0.001, rho=0.99)<\/code><\/pre>\n<p>RMSprop\u4e5f\u662f\u4e00\u79cd\u6bd4\u8f83\u5e38\u7528\u7684\u4f18\u5316\u5668,\u5b83\u901a\u8fc7\u8ba1\u7b97\u68af\u5ea6\u7684\u5747\u65b9\u6839\u6765\u8c03\u6574\u5b66\u4e60\u7387,\u53ef\u4ee5\u52a0\u901fSGD\u5728\u4e00\u5b9a\u7a0b\u5ea6\u4e0a\u7684\u6536\u655b\u3002<\/p>\n<ol start=\"4\">\n<li>Adadelta\u4f18\u5316\u5668:<\/li>\n<\/ol>\n<pre><code>python\n# PyTorch\noptimizer = torch.optim.Adadelta(model.parameters(), lr=1.0)\n\n# TensorFlow \noptimizer = tf.keras.optimizers.Adadelta(learning_rate=1.0)<\/code><\/pre>\n<p>Adadelta\u662f\u4e00\u4e2a\u81ea\u9002\u5e94\u5b66\u4e60\u7387\u65b9\u6cd5,\u5b83\u4e0d\u9700\u8981\u8bbe\u7f6e\u9ed8\u8ba4\u5b66\u4e60\u7387,\u4f1a\u6839\u636e\u53c2\u6570\u7684\u66f4\u65b0\u5e45\u5ea6\u81ea\u52a8\u8c03\u6574\u5b66\u4e60\u7387\u3002<\/p>\n<\/li>\n<li>\n<p>\u5b9a\u4e49\u8bc4\u4f30\u6307\u6807(metrics):\u7528\u4e8e\u8bc4\u4f30\u6a21\u578b\u6548\u679c,\u5e38\u7528\u7684\u6709\u51c6\u786e\u7387accuracy\u3001\u7cbe\u786e\u7387precision\u3001\u53ec\u56de\u7387recall\u7b49\u3002\u8fd9\u91cc\u7ed9\u51fa\u8be6\u7ec6\u7684\u4ee3\u7801\u5b9e\u73b0:<\/p>\n<\/li>\n<\/ol>\n<pre><code>python\n# \u5bfc\u5165\u635f\u5931\u51fd\u6570\u548c\u4f18\u5316\u5668\nfrom tensorflow.keras.losses import CategoricalCrossentropy\nfrom tensorflow.keras.optimizers import Adam\n\n# \u7f16\u8bd1\u6a21\u578b\nmodel.compile(\n    # \u9009\u62e9\u4ea4\u53c9\u71b5\u635f\u5931\u51fd\u6570,\u56e0\u4e3a\u6211\u4eec\u662f\u591a\u5206\u7c7b\u4efb\u52a1\n    loss=CategoricalCrossentropy(),\n\n    # \u9009\u62e9Adam\u4f18\u5316\u5668\n    optimizer=Adam(learning_rate=0.001), \n\n    # \u6d4b\u91cf\u51c6\u786e\u7387\u548c\u7cbe\u786e\u7387\n    metrics=[\n        &#039;accuracy&#039;, \n        tf.keras.metrics.Precision(),\n    ]\n)<\/code><\/pre>\n<p>- \u6211\u4eec\u9009\u62e9CategoricalCrossentropy\u4f5c\u4e3a\u635f\u5931\u51fd\u6570,\u56e0\u4e3a\u8fd9\u4e2a\u662f\u4e00\u4e2a\u591a\u5206\u7c7b\u7684CRNN\u6a21\u578b,CategoricalCrossentropy\u9002\u7528\u4e8e\u591a\u5206\u7c7b\u3002- \u9009\u62e9Adam\u4f18\u5316\u5668,\u5b66\u4e60\u7387\u4e3a0.001\u3002<\/p>\n<p>Adam\u662f\u6bd4\u8f83\u5e38\u7528\u7684\u4f18\u5316\u5668,needs tuning\u7684\u5b66\u4e60\u7387\u3002<\/p>\n<ul>\n<li>\n<p>\u6211\u4eec\u5b9a\u4e49\u4e24\u4e2a\u8bc4\u4f30\u6307\u6807:accuracy\u548cprecision\u3002<br \/>\n- accuracy:\u51c6\u786e\u7387,\u8861\u91cf\u6b63\u786e\u9884\u6d4b\u7684\u6bd4\u4f8b\u3002<br \/>\n- precision:\u7cbe\u786e\u7387,\u8861\u91cf\u88ab\u6a21\u578b\u5224\u65ad\u4e3a\u6b63\u6837\u672c\u4e2d\u5b9e\u9645\u6b63\u6837\u672c\u7684\u6bd4\u4f8b\u3002<\/p>\n<p>\u6240\u4ee5\u7f16\u8bd1\u6a21\u578b\u4e3b\u8981 set 3\u4e2a\u65b9\u9762:<\/p>\n<ol>\n<li>\u635f\u5931\u51fd\u6570:\u8861\u91cf\u9884\u6d4b\u548c\u771f\u5b9e\u6807\u7b7e\u5dee\u5f02\u7684\u5ea6\u91cf\u3002<\/li>\n<\/ol>\n<\/li>\n<\/ul>\n<p>\u200b         2.\u4f18\u5316\u5668:\u66f4\u65b0\u6a21\u578b\u53c2\u6570\u4ee5\u6700\u5c0f\u5316\u635f\u5931\u51fd\u6570\u3002<\/p>\n<p>\u200b       3.\u8bc4\u4f30\u6307\u6807:\u7528\u4e8e\u8bc4\u4ef7\u6a21\u578b\u7684\u6548\u679c\u3002\u9009\u62e9\u5408\u9002\u7684\u635f\u5931\u51fd\u6570\u3001\u4f18\u5316\u5668\u548c\u8bc4\u4f30\u6307\u6807\u5bf9\u6a21\u578b\u6027\u80fd\u4f1a\u6709\u6bd4\u8f83\u5927\u7684\u5f71\u54cd,\u4e5f\u662f\u6a21\u578b\u8c03\u4f18\u7684\u4e00\u4e2a\u91cd\u8981\u6b65\u9aa4\u3002<\/p>\n<p>\u200b               <font color=red>\u8bad\u7ec3\u6a21\u578b:\u8c03\u7528model.fit\u8fdb\u884c\u8bad\u7ec3,\u8f93\u5165\u7279\u5f81\u548c\u6807\u7b7e\u3002<\/font><\/p>\n<p>\u200b                 \u6211\u4eec\u5728\u8bad\u7ec3\u6a21\u578b\u7684\u65f6\u5019\u4f7f\u7528\u4f18\u5316\u5668\u4e0d\u65ad\u66f4\u65b0\u6a21\u578b\u53c2\u6570,\u4f7f\u635f\u5931\u51fd\u6570\u6700\u5c0f\u5316\u7684\u8fc7\u7a0b\u3002\ud83d\udc47\u6211\u4f1a\u4f7f\u7528TensorFlow\u4e3a\u4f8b\u7ed9\u51fa\u8bad\u7ec3\u4ee3\u7801\u5b9e\u73b0:<\/p>\n<pre><code>python\n# \u51c6\u5907\u8bad\u7ec3\u6570\u636e\ntrain_dataset = ...\n\n# \u7f16\u8bd1\u6a21\u578b\nmodel.compile(loss=..., optimizer=..., metrics=...)\n\n# \u8bad\u7ec3\u6a21\u578b\nmodel.fit(\n    train_dataset,\n    epochs=10,        # \u8fed\u4ee3\u6b21\u6570\n    batch_size=64,    # \u6279\u5927\u5c0f\n    validation_split=0.2 #  validation\u6bd4\u4f8b \n)<\/code><\/pre>\n<p>model.fit\u9700\u8981\u8f93\u5165:<\/p>\n<ul>\n<li>\n<p>train_dataset:\u8bad\u7ec3\u6570\u636e,\u53ef\u4ee5\u662fNumpy\u6570\u7ec4,TensorFlow Dataset\u7b49\u3002<\/p>\n<\/li>\n<li>\n<p>epochs:\u8fed\u4ee3\u6b21\u6570,\u5168\u90e8\u8bad\u7ec3\u6570\u636e\u4f1a\u88ab\u8fed\u4ee3\u4f7f\u7528epochs\u6b21\u3002<\/p>\n<\/li>\n<li>\n<p>batch_size:\u6bcf\u4e2abatch\u4f7f\u7528\u7684\u6837\u672c\u6570\u91cf\u3002\u5982\u679c\u8bbe\u7f6e\u4e3aNone,\u90a3\u4e48\u53d6\u8bad\u7ec3\u6570\u636e\u4e2d\u7684\u5168\u90e8\u6837\u672c\u3002<\/p>\n<\/li>\n<li>\n<p>validation_split:\u4ece\u8bad\u7ec3\u6570\u636e\u4e2d\u5206\u5272\u7684\u6bd4\u4f8b\u4f5c\u4e3a\u9a8c\u8bc1\u96c6\u3002<\/p>\n<p>\u4e3b\u8981\u7684\u8bad\u7ec3\u8fc7\u7a0b\u662f:<\/p>\n<ol>\n<li>\u4ece\u8bad\u7ec3\u6570\u636e\u4e2d\u968f\u673a\u53d6\u4e00\u4e2abatch\u7684\u6570\u636e\u3002<\/li>\n<li>\u4f7f\u7528\u4f18\u5316\u5668\u548c\u635f\u5931\u51fd\u6570\u8ba1\u7b97\u5728\u8be5batch\u6570\u636e\u4e0a\u7684\u68af\u5ea6\u548c\u635f\u5931\u3002<\/li>\n<li>\u4f7f\u7528\u68af\u5ea6\u66f4\u65b0\u6a21\u578b\u53c2\u6570\u3002<\/li>\n<li>\u91cd\u590d1-3\u6b65\u9aa4,\u76f4\u5230\u5b8c\u6210\u4e00\u4e2aepoch\u3002<\/li>\n<li>\u5982\u679c\u8bbe\u7f6e\u4e86\u9a8c\u8bc1\u96c6,\u5728\u6bcf\u4e2aepoch\u7ed3\u675f\u65f6\u4f1a\u8bc4\u4f30\u9a8c\u8bc1\u96c6\u7684\u635f\u5931\u548c\u6307\u6807\u3002<\/li>\n<li>\u91cd\u590d1-5\u6b65\u9aa4,\u76f4\u5230\u5b8c\u6210\u6240\u6709\u7684epochs\u3002<\/li>\n<li>\u4e00\u4e9b\u9700\u8981\u6ce8\u610f\u7684\u70b9:\n<ol>\n<li>batch_size\u4e0d\u8981\u592a\u5927\u6216\u592a\u5c0f,\u4ee5\u514d\u5f71\u54cd\u6a21\u578b\u8bad\u7ec3\u6548\u679c\u3002<\/li>\n<li>learning_rate\u7684\u8bbe\u7f6e\u4f1a\u76f4\u63a5\u5f71\u54cd\u6a21\u578b\u7684\u6536\u655b\u901f\u5ea6\u548c\u6548\u679c,\u9700\u8981\u6839\u636e\u5b9e\u9645\u6570\u636e\u8fdb\u884c\u9009\u62e9\u3002<\/li>\n<li>\u66f4\u591a\u7684epochs\u5e76\u4e0d\u4e00\u5b9a\u4f1a\u5e26\u6765\u66f4\u597d\u7684\u6548\u679c,\u9700\u8981\u65e9\u505c\u4ee5\u9632\u6b62\u8fc7\u62df\u5408\u3002<\/li>\n<li>\u53ef\u4ee5\u6839\u636e\u9a8c\u8bc1\u96c6\u7684\u6548\u679c\u9009\u62e9\u6700\u4f18\u7684\u4e00\u8f6e\u4f5c\u4e3a\u6a21\u578b\u7684\u6700\u7ec8\u53c2\u6570\u3002<\/li>\n<\/ol>\n<\/li>\n<\/ol>\n<\/li>\n<\/ul>\n<p>\u200b               <font color=red>\u8bc4\u4f30\u6a21\u578b:\u8c03\u7528model.evaluate\u8fdb\u884c\u8bc4\u4f30\u3002<\/font><\/p>\n<p>\u200b                   \u6211\u4eec\u8981\u660e\u767d\ud83d\ude00\u8bc4\u4f30\u6a21\u578b\u5c31\u662f\u4f7f\u7528\u9a8c\u8bc1\u96c6\u6216\u6d4b\u8bd5\u96c6\u7684\u6570\u636e\u8bc4\u4f30\u6a21\u578b\u7684\u6548\u679c,\u4e3b\u8981\u662f\u8bc4\u4f30\u635f\u5931\u51fd\u6570\u548c\u4e4b\u524d\u5b9a\u4e49\u7684\u8bc4\u4f30\u6307\u6807\u3002\u4e0b\u9762\u7ed9\u4e00\u4e0b\u9a9a\u64cd\u4f5c\u4ee3\u7801\u5b9e\u73b0\u5982\u4e0b:<\/p>\n<pre><code>python\n# \u51c6\u5907\u9a8c\u8bc1\u6570\u636e \nval_dataset = ...\n\n# \u8bc4\u4f30\u6a21\u578b  \nresults = model.evaluate(val_dataset, batch_size=64)<\/code><\/pre>\n<p>model.evaluate\u4f1a\u8fd4\u56de\u4e00\u4e2a\u7ed3\u679c\u5217\u8868,\u7b2c\u4e00\u4e2a\u5143\u7d20\u662f\u635f\u5931\u51fd\u6570\u7684\u503c,\u4e4b\u540e\u7684\u5143\u7d20\u662f\u4e4b\u524d\u5b9a\u4e49\u7684\u8bc4\u4f30\u6307\u6807\u7684\u503c\u3002<\/p>\n<p>\u6bd4\u5982,\u5982\u679c\u4e4b\u524d\u5b9a\u4e49\u7684\u6307\u6807\u662faccuracy\u548cprecision,\u90a3\u4e48results\u53ef\u4ee5\u662f:<br \/>\n[0.32, 0.85, 0.70]<\/p>\n<p>\u5176\u4e2d,<br \/>\nresults[0]\u662f\u635f\u5931\u51fd\u6570\u503c<br \/>\nresults[1]\u662faccuracy<br \/>\nresults[2]\u662fprecision<\/p>\n<p>\u8bc4\u4f30\u6a21\u578b\u4e3b\u8981\u662f\u4e3a\u4e86:<\/p>\n<ol>\n<li>\n<p>\u9009\u62e9\u6700\u4f18\u6a21\u578b:\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d,\u53ef\u4ee5\u6839\u636e\u9a8c\u8bc1\u96c6\u8bc4\u4f30\u7ed3\u679c\u9009\u62e9\u6700\u4f18\u7684\u4e00\u8f6e\u6a21\u578b\u53c2\u6570\u3002<\/p>\n<\/li>\n<li>\n<p>\u68c0\u9a8c\u6cdb\u5316\u80fd\u529b:\u4f7f\u7528\u6d4b\u8bd5\u96c6\u8bc4\u4f30\u6a21\u578b,\u5224\u65ad\u662f\u5426\u8fc7\u62df\u5408\u4ee5\u53ca\u6a21\u578b\u7684\u6cdb\u5316\u80fd\u529b\u3002<\/p>\n<\/li>\n<li>\n<p>\u8c03\u4f18\u6a21\u578b:\u6839\u636e\u8bc4\u4f30\u7ed3\u679c\u5224\u65ad\u6a21\u578b\u7684\u6548\u679c,\u5e76\u6839\u636e\u7ed3\u679c\u8c03\u6574\u6a21\u578b\u7ed3\u6784\u3001\u8bad\u7ec3\u53c2\u6570\u7b49\u4ee5\u6539\u8fdb\u6a21\u578b\u3002<\/p>\n<\/li>\n<li>\n<p>\u9884\u6d4b\u65b0\u6570\u636e\u7684\u6548\u679c:\u6211\u4eec\u5728\u6700\u7ec8\u9009\u62e9\u7684\u6700\u4f18\u6a21\u578b\u4e0a\u8fdb\u884c\u8bc4\u4f30,\u5f97\u5230\u7684\u7ed3\u679c\u53ef\u4ee5\u4f5c\u4e3a\u5bf9\u65b0\u6570\u636e\u8fdb\u884c\u9884\u6d4b\u7684\u4e00\u4e2a\u53c2\u8003\u3002<\/p>\n<p>\u6240\u4ee5,\u8bc4\u4f30\u6a21\u578b\u662f\u8861\u91cf\u6a21\u578b\u6548\u679c\u548c\u9009\u62e9\u6700\u4f18\u6a21\u578b\u7684\u91cd\u8981\u6b65\u9aa4\u3002<\/p>\n<p>\u5728\u8bad\u7ec3\u548c\u8c03\u4f18\u6a21\u578b\u7684\u8fc7\u7a0b\u4e2d,\u9700\u8981\u4e0d\u65ad\u6839\u636e\u9a8c\u8bc1\u96c6\u548c\u6d4b\u8bd5\u96c6\u7684\u8bc4\u4f30\u7ed3\u679c\u8fdb\u884c\u5224\u65ad\u548c\u6539\u8fdb\u3002<\/p>\n<p>\u800c\u5728\u6700\u7ec8\u786e\u5b9a\u6700\u4f18\u6a21\u578b\u540e,\u4f7f\u7528\u6d4b\u8bd5\u96c6\u8fdb\u884c\u8bc4\u4f30\u53ef\u4ee5\u4e86\u89e3\u6a21\u578b\u7684\u6cdb\u5316\u80fd\u529b\u4ee5\u53ca\u5bf9\u65b0\u6570\u636e\u7684\u9884\u6d4b\u6548\u679c\u3002<\/p>\n<\/li>\n<\/ol>\n<p>\u200b               <font color=red>\u9884\u6d4b\u65b0\u6570\u636e:\u8c03\u7528model.predict\u8fdb\u884c\u9884\u6d4b\u3002<\/font><\/p>\n<p>CRNN\u6a21\u578b\u7684\u6ce8\u610f\u4e8b\u9879- \u9009\u62e9\u5408\u9002\u7684\u5377\u79ef\u6838\u5927\u5c0f\u548c\u6570\u91cf,\u6c60\u5316\u53c2\u6570\u3002<\/p>\n<p>\u8fc7\u6d45\u4f1a\u63d0\u53d6\u4e0d\u5230\u597d\u7279\u5f81,\u8fc7\u6df1\u4f1a\u9020\u6210\u6570\u636e\u548c\u8ba1\u7b97\u91cf\u7206\u70b8\u3002<\/p>\n<p>LSTM\u5c42\u7684\u53c2\u6570\u9009\u62e9\u6bd4\u8f83\u96be,\u9700\u8981\u6839\u636e\u4efb\u52a1\u7684\u590d\u6742\u5ea6\u9009\u62e9\u3002<\/p>\n<p>\u592a\u7b80\u5355\u4f1a\u9020\u6210\u6a21\u578b\u8868\u8fbe\u80fd\u529b\u4e0d\u8db3,\u592a\u590d\u6742\u4f1a\u5bfc\u81f4\u6cdb\u5316\u80fd\u529b\u4e0b\u964d\u548c\u8ba1\u7b97\u5f00\u9500\u589e\u5927\u3002<\/p>\n<p>\u5145\u5206\u7684\u6570\u636e\u548c\u6807\u7b7e\u5f88\u91cd\u8981\u3002\u5e8f\u5217\u6570\u636e\u4e00\u822c\u6bd4\u8f83\u96be\u83b7\u5f97,\u8fd9\u4f1a\u76f4\u63a5\u5f71\u54cd\u6a21\u578b\u7684\u6548\u679c\u3002<\/p>\n<p>\u8fc7\u62df\u5408\u662f\u4e00\u4e2a\u9700\u8981\u5173\u6ce8\u7684\u95ee\u9898\u3002\u53ef\u4ee5\u901a\u8fc7 dropout\u3001\u6570\u636e\u589e\u5f3a\u3001L2\u6b63\u5219\u5316\u7b49\u65b9\u6cd5\u8fdb\u884c\u7f13\u89e3\u3002<\/p>\n<p>\u9009\u62e9\u6070\u5f53\u7684\u8bad\u7ec3\u53c2\u6570,\u5982\u5b66\u4e60\u7387\u3001\u8fed\u4ee3\u6b21\u6570\u7b49\u4f1a\u5bf9\u6a21\u578b\u7684\u6548\u679c\u4ea7\u751f\u6bd4\u8f83\u5927\u5f71\u54cd,\u9700\u8981\u8fdb\u884c\u8c03\u53c2\u3002<\/p>\n<p>CRNN\u6a21\u578b\u662f\u4e00\u79cd\u5e8f\u5217\u5230\u5e8f\u5217(Sequence to Sequence)\u7684\u6a21\u578b,\u7528\u4e8e\u5904\u7406\u5e8f\u5217\u6570\u636e,\u6bd4\u5982\u6587\u672c,\u8bed\u97f3\u7b49\u3002<\/p>\n<p>\u5b83\u901a\u5e38\u5305\u542b\u4ee5\u4e0b\u51e0\u4e2a\u90e8\u5206:<\/p>\n<ol>\n<li>\n<p>\u5377\u79ef\u5c42(Convolutional Layer):\u7528\u4e8e\u7279\u5f81\u63d0\u53d6,\u53ef\u4ee5\u4f7f\u7528\u591a\u5c42\u5377\u79ef\u63d0\u53d6\u4e0d\u540c\u7ea7\u522b\u7684\u7279\u5f81\u3002<\/p>\n<\/li>\n<li>\n<p>\u6c60\u5316\u5c42(Pooling Layer):\u7528\u4e8e\u51cf\u5c11\u7279\u5f81\u7ef4\u5ea6,\u964d\u4f4e\u8ba1\u7b97\u590d\u6742\u5ea6\u3002<\/p>\n<\/li>\n<li>\n<p>\u91cd\u590d\u5c42(Recurrent Layer):\u901a\u5e38\u662fLSTM\u6216GRU,\u7528\u4e8e capturing long-range dependencies\u3002<\/p>\n<\/li>\n<li>\n<p>\u5168\u8fde\u63a5\u5c42(Fully Connected Layer):\u7528\u4e8e\u6700\u7ec8\u7684\u9884\u6d4b\u6216\u5206\u7c7b\u3002<\/p>\n<p>\u4e00\u4e2a\u7b80\u5355\u7684CRNN\u6a21\u578b\u53ef\u4ee5\u5982\u4e0b\u5b9e\u73b0:<\/p>\n<\/li>\n<\/ol>\n<pre><code>java\npublic class CRNNModel {\n    \/\/ \u5377\u79ef\u5c42\n    private ConvolutionalLayer convLayer1;\n    private ConvolutionalLayer convLayer2;\n    \/\/ \u6c60\u5316\u5c42\n    private PoolingLayer poolLayer1;\n    private PoolingLayer poolLayer2;\n    \/\/ \u9012\u5f52\u5c42\n    private LSTMLayer lstmLayer;\n    \/\/ \u5168\u8fde\u63a5\u5c42\n    private FullyConnectedLayer fcLayer;\n\n    public CRNNModel() {\n        \/\/ \u5377\u79ef\u5c42\n        convLayer1 = new ConvolutionalLayer(3, 3, 16); \n        convLayer2 = new ConvolutionalLayer(3, 3, 32);\n        \/\/ \u6c60\u5316\u5c42\n        poolLayer1 = new MaxPoolLayer(2, 2);\n        poolLayer2 = new MaxPoolLayer(2, 2);\n        \/\/ LSTM\u5c42\n        lstmLayer = new LSTMLayer(32, 64);\n        \/\/ \u5168\u8fde\u63a5\u5c42\n        fcLayer = new FullyConnectedLayer(64, 10);\n    }\n\n    public void train(int[][] inputs, int[] targets) {\n        \/\/ \u5377\u79ef\n        int[][] conv1Out = convLayer1.forwardPropagate(inputs);\n        int[][] pool1Out = poolLayer1.forwardPropagate(conv1Out);\n        int[][] conv2Out = convLayer2.forwardPropagate(pool1Out);\n        int[][] pool2Out = poolLayer2.forwardPropagate(conv2Out);\n\n        \/\/ LSTM\n        int[][] lstmOut = lstmLayer.forwardPropagate(pool2Out);\n\n        \/\/ \u5168\u8fde\u63a5\n        int[] fcOut = fcLayer.forwardPropagate(lstmOut);\n\n        \/\/ \u8ba1\u7b97\u8bef\u5dee\n        int[] error = LossFunction.crossEntropyError(fcOut, targets);\n\n        \/\/ \u8bef\u5dee\u56de\u4f20\n        fcLayer.backwardPropagate(error);\n        lstmLayer.backwardPropagate(fcLayer.getErrors());\n        ...\n\n    }\n}<\/code><\/pre>\n<p>CRNN\u6a21\u578b\u4e3b\u8981\u7528\u4e8e\u5e8f\u5217\u6570\u636e\u7684\u5206\u7c7b\u548c\u9884\u6d4b,\u6bd4\u5982:<\/p>\n<ul>\n<li>\u6587\u672c\u5206\u7c7b:\u65b0\u95fb\u5206\u7c7b,\u8bc4\u8bba\u5206\u7c7b\u7b49<br \/>\n- \u8bed\u97f3\u8bc6\u522b:\u8bed\u97f3\u8f6c\u6587\u5b57<br \/>\n- \u624b\u5199\u8bc6\u522b:\u624b\u5199\u6570\u5b57\u6216\u6587\u5b57\u8bc6\u522b<br \/>\n- \u65f6\u95f4\u5e8f\u5217\u9884\u6d4b:\u80a1\u4ef7\u9884\u6d4b,\u9500\u91cf\u9884\u6d4b\u7b49\u6240\u4ee5CRNN\u6a21\u578b\u662f\u4e00\u4e2a\u529f\u80fd\u5f88\u5f3a\u5927\u7684\u5e8f\u5217\u5efa\u6a21\u6a21\u578b,\u5728\u8bb8\u591a\u5e8f\u5217\u6570\u636e\u5904\u7406\u4efb\u52a1\u4e2d\u90fd\u6709\u5f88\u597d\u7684\u5e94\u7528\u3002<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u524d\u8a00 \u6700\u8fd1\u9879\u76ee\u4e2d\u8981\u4f7f\u7528\u8bed\u97f3\u8bc6\u522b\uff0c\u67e5\u9605\u8d44\u6599\u627e\u5230\u4e86\u4e00\u4e2a\u795e\u5947\u7684\u6a21\u578b\uff1aCRNN \uff0c\u901a\u8fc7\u7f16\u5199\u6587\u7ae0\uff0c\u53ef\u4ee5\u6559\u60a8\u5927\u6982\u77e5\u9053\u8fd9\u4e2aC [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"om_disable_all_campaigns":false,"_mi_skip_tracking":false},"categories":[1],"tags":[],"aioseo_notices":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v20.5 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09 - \u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug<\/title>\n<meta name=\"description\" content=\"The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff08\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff09\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09 - \u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug\" \/>\n<meta property=\"og:description\" content=\"The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff08\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff09\" \/>\n<meta property=\"og:url\" content=\"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/\" \/>\n<meta property=\"og:site_name\" content=\"\u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug\" \/>\n<meta property=\"article:published_time\" content=\"2023-05-08T02:52:28+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-05-08T02:54:11+00:00\" \/>\n<meta name=\"author\" content=\"\u603b\u662f\u5e78\u798f\u7684\u8001\u8c4c\u8c46\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"\u603b\u662f\u5e78\u798f\u7684\u8001\u8c4c\u8c46\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"3\u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/\",\"url\":\"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/\",\"name\":\"\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09 - \u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug\",\"isPartOf\":{\"@id\":\"https:\/\/zhaocunwei.co.uk\/#website\"},\"datePublished\":\"2023-05-08T02:52:28+00:00\",\"dateModified\":\"2023-05-08T02:54:11+00:00\",\"author\":{\"@id\":\"https:\/\/zhaocunwei.co.uk\/#\/schema\/person\/dfb1dc0fc4a330c41908d477cd99c0b4\"},\"description\":\"The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff08\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff09\",\"breadcrumb\":{\"@id\":\"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/#breadcrumb\"},\"inLanguage\":\"zh-CN\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/zhaocunwei.co.uk\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/zhaocunwei.co.uk\/#website\",\"url\":\"https:\/\/zhaocunwei.co.uk\/\",\"name\":\"\u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug\",\"description\":\"This site shares programming bugs\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/zhaocunwei.co.uk\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-CN\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/zhaocunwei.co.uk\/#\/schema\/person\/dfb1dc0fc4a330c41908d477cd99c0b4\",\"name\":\"\u603b\u662f\u5e78\u798f\u7684\u8001\u8c4c\u8c46\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-CN\",\"@id\":\"https:\/\/zhaocunwei.co.uk\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/secure.gravatar.com\/avatar\/4226cc1ca6640507df1d2d4ba3da7a62?s=96&d=mm&r=g\",\"contentUrl\":\"https:\/\/secure.gravatar.com\/avatar\/4226cc1ca6640507df1d2d4ba3da7a62?s=96&d=mm&r=g\",\"caption\":\"\u603b\u662f\u5e78\u798f\u7684\u8001\u8c4c\u8c46\"},\"sameAs\":[\"http:\/\/zhaocunwei.co.uk\"],\"url\":\"https:\/\/zhaocunwei.co.uk\/index.php\/author\/18500103508163-com\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09 - \u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug","description":"The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff08\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff09","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/","og_locale":"zh_CN","og_type":"article","og_title":"\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09 - \u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug","og_description":"The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff08\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff09","og_url":"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/","og_site_name":"\u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug","article_published_time":"2023-05-08T02:52:28+00:00","article_modified_time":"2023-05-08T02:54:11+00:00","author":"\u603b\u662f\u5e78\u798f\u7684\u8001\u8c4c\u8c46","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"\u603b\u662f\u5e78\u798f\u7684\u8001\u8c4c\u8c46","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"3\u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/","url":"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/","name":"\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09 - \u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug","isPartOf":{"@id":"https:\/\/zhaocunwei.co.uk\/#website"},"datePublished":"2023-05-08T02:52:28+00:00","dateModified":"2023-05-08T02:54:11+00:00","author":{"@id":"https:\/\/zhaocunwei.co.uk\/#\/schema\/person\/dfb1dc0fc4a330c41908d477cd99c0b4"},"description":"The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff08\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff09","breadcrumb":{"@id":"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/#breadcrumb"},"inLanguage":"zh-CN","potentialAction":[{"@type":"ReadAction","target":["https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/zhaocunwei.co.uk\/index.php\/2023\/05\/08\/the-project-uses-voice-interaction\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/zhaocunwei.co.uk\/"},{"@type":"ListItem","position":2,"name":"\u9879\u76ee\u4f7f\u7528\u8bed\u97f3\u4ea4\u4e92\uff0c\u987a\u5e26\u719f\u6089\u4e86\u4e00\u4e0bCRNN\uff0c\u6211\u60f3\u901a\u8fc7\u6587\u7ae0\u63cf\u8ff0\u4ecb\u7ecd\u4e00\u4e0bCRNN\u7684\u5177\u4f53\u7528\u6cd5\uff08The project uses voice interaction, and I am familiar with CRNN by the way. I would like to introduce the specific usage of CRNN through the article description\uff09"}]},{"@type":"WebSite","@id":"https:\/\/zhaocunwei.co.uk\/#website","url":"https:\/\/zhaocunwei.co.uk\/","name":"\u672c\u7f51\u7ad9\u5206\u4eab\u7f16\u7a0b\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684bug","description":"This site shares programming bugs","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/zhaocunwei.co.uk\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-CN"},{"@type":"Person","@id":"https:\/\/zhaocunwei.co.uk\/#\/schema\/person\/dfb1dc0fc4a330c41908d477cd99c0b4","name":"\u603b\u662f\u5e78\u798f\u7684\u8001\u8c4c\u8c46","image":{"@type":"ImageObject","inLanguage":"zh-CN","@id":"https:\/\/zhaocunwei.co.uk\/#\/schema\/person\/image\/","url":"https:\/\/secure.gravatar.com\/avatar\/4226cc1ca6640507df1d2d4ba3da7a62?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/4226cc1ca6640507df1d2d4ba3da7a62?s=96&d=mm&r=g","caption":"\u603b\u662f\u5e78\u798f\u7684\u8001\u8c4c\u8c46"},"sameAs":["http:\/\/zhaocunwei.co.uk"],"url":"https:\/\/zhaocunwei.co.uk\/index.php\/author\/18500103508163-com\/"}]}},"_links":{"self":[{"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/posts\/486"}],"collection":[{"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/comments?post=486"}],"version-history":[{"count":1,"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/posts\/486\/revisions"}],"predecessor-version":[{"id":487,"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/posts\/486\/revisions\/487"}],"wp:attachment":[{"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/media?parent=486"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/categories?post=486"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/zhaocunwei.co.uk\/index.php\/wp-json\/wp\/v2\/tags?post=486"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}