don't forget to preprocessing, it really important

preprocessing.py

@@ -0,0 +1,105 @@
+import os
+import numpy as np
+from pydub import AudioSegment
+from scipy.io import wavfile
+
+def trim_wav(input_file, output_file, target_duration_minutes=24, target_duration_seconds=10):
+    """
+    Memotong file WAV sesuai durasi target
+    """
+    try:
+        # Load file WAV
+        audio = AudioSegment.from_wav(input_file)
+        
+        # Konversi target durasi ke milidetik
+        target_duration_ms = (target_duration_minutes * 60 + target_duration_seconds) * 1000
+        
+        # Cek apakah durasi audio melebihi target
+        if len(audio) <= target_duration_ms:
+            print(f"File {input_file} memiliki durasi kurang dari target, diloncati")
+            return None
+            
+        # Potong audio sesuai durasi yang diinginkan
+        trimmed_audio = audio[:target_duration_ms]
+        
+        # Export file yang sudah dipotong
+        trimmed_audio.export(output_file, format="wav")
+        
+        print(f"File berhasil dipotong menjadi {target_duration_minutes} menit {target_duration_seconds} detik")
+        return output_file
+        
+    except Exception as e:
+        print(f"Terjadi kesalahan saat memotong file: {str(e)}")
+        return None
+
+def segment_wav_file(input_file, output_dir, segment_duration=5, overlap_percentage=0.5):
+    """
+    Memotong file WAV menjadi segmen-segmen dengan overlap
+    """
+    # Create output directory if not exists
+    os.makedirs(output_dir, exist_ok=True)
+    
+    # Read the input WAV file
+    sample_rate, audio_data = wavfile.read(input_file)
+    
+    # Convert segment duration to samples
+    samples_per_segment = int(segment_duration * sample_rate)
+    samples_overlap = int(samples_per_segment * overlap_percentage)
+    
+    # Calculate step size between segments
+    step_size = samples_per_segment - samples_overlap
+    
+    # Segment the audio
+    segmented_files = []
+    for start in range(0, len(audio_data) - samples_per_segment + 1, step_size):
+        end = start + samples_per_segment
+        segment = audio_data[start:end]
+        
+        # Generate output filename dengan format "sound_segment_X.wav"
+        segment_filename = f"ridho_segment_{start//step_size}.wav"
+        output_path = os.path.join(output_dir, segment_filename)
+        
+        # Write segmented audio to file
+        wavfile.write(output_path, sample_rate, segment)
+        segmented_files.append(output_path)
+    
+    print(f"Berhasil membuat {len(segmented_files)} segmen")
+    return segmented_files
+
+def process_audio_pipeline(input_wav_file, output_dir):
+    """
+    Fungsi utama untuk memproses single file WAV
+    """
+    try:
+        # Buat direktori output jika belum ada
+        os.makedirs(output_dir, exist_ok=True)
+        
+        # Step 1: Trim to 24:10
+        trimmed_wav = os.path.join(output_dir, "trimmed.wav")
+        trimmed_result = trim_wav(input_wav_file, trimmed_wav)
+        if not trimmed_result:
+            return
+            
+        # Step 2: Create segments
+        segments_dir = os.path.join(output_dir, "segments")
+        segmented_files = segment_wav_file(
+            trimmed_wav,
+            segments_dir,
+            segment_duration=5,
+            overlap_percentage=0.5
+        )
+        
+        print("Proses selesai!")
+        return segmented_files
+        
+    except Exception as e:
+        print(f"Terjadi kesalahan dalam pipeline: {str(e)}")
+        return None
+
+# Contoh penggunaan
+if __name__ == "__main__":
+    # Sesuaikan path file input dan direktori output
+    input_wav = "full recording.wav"
+    output_dir = "/nama/path/to/target/dataset"
+    
+    process_audio_pipeline(input_wav, output_dir)