Modified record_analysis.py

IT20201364/record_analysis.py

+import sys
+import os
+import sounddevice as sd
+import numpy as np
+import librosa
+import keras.models
+
+# Load the saved emotion identification model
+loaded_model = keras.models.load_model('C:/Users/dell/Desktop/F_Project/Virtual_Assistant/my_model.h5')
+classLabels = ('Angry', 'Fear', 'Disgust', 'Happy', 'Sad', 'Surprised', 'Neutral')
+
+def record_and_analyze_user_input(chatbot):
+    duration = 2.5  # Set the duration of the recording (in seconds)
+    sample_rate = 22050 * 2  # Set the sample rate of the recording
+
+    # Redirect standard output to a null device
+    sys.stdout = open(os.devnull, 'w')
+
+    # Record the live audio
+    recording = sd.rec(int(duration * sample_rate), samplerate=sample_rate, channels=1)
+    sd.wait()  # Wait until the recording is finished
+
+    # Restore standard output
+    sys.stdout = sys.__stdout__
+
+    # Preprocess the recorded audio
+    mfccs = librosa.feature.mfcc(y=recording.flatten(), sr=sample_rate, n_mfcc=39)
+    input_data = mfccs[np.newaxis, ..., np.newaxis]
+
+    # Perform emotion prediction
+    with np.printoptions(suppress=True):
+        predictions = loaded_model.predict(input_data)
+
+    # Get the predicted label
+    predicted_label = classLabels[np.argmax(predictions)]
+
+    # Store the predicted emotion in the array
+    chatbot.sentiment_analysis.append(predicted_label)
+
+    return predicted_label
\ No newline at end of file