Create app.py

app.py

@@ -0,0 +1,93 @@
+import gradio as gr
+import numpy as np
+import librosa
+from scipy.signal import find_peaks
+from sklearn.cluster import KMeans
+
+def decodificar(audio):
+
+    if audio is None:
+        return "No audio"
+
+    path = audio
+
+    y, sr = librosa.load(path, sr=None)
+
+    frame = int(sr * 0.04)
+
+    stft = np.abs(
+        librosa.stft(
+            y,
+            n_fft=frame*2,
+            hop_length=frame
+        )
+    )
+
+    freqs = librosa.fft_frequencies(sr=sr)
+
+    tonos = []
+
+    for f in stft.T:
+
+        if np.max(f) == 0:
+            continue
+
+        f = f / np.max(f)
+
+        peaks, _ = find_peaks(f, height=0.2)
+
+        if len(peaks) == 0:
+            continue
+
+        peak_freqs = freqs[peaks]
+
+        peak_freqs = peak_freqs[
+            (peak_freqs > 300) &
+            (peak_freqs < 4000)
+        ]
+
+        if len(peak_freqs):
+            tonos.append(peak_freqs[0])
+
+    if len(tonos) < 10:
+        return "Sin señal tonal clara"
+
+    tonos = np.array(tonos).reshape(-1,1)
+
+    kmeans = KMeans(n_clusters=12, n_init=10)
+    kmeans.fit(tonos)
+
+    centros = sorted(kmeans.cluster_centers_.flatten())
+
+    letras = "ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"
+
+    texto=""
+
+    for f in tonos.flatten():
+
+        cercano = min(centros, key=lambda x: abs(x-f))
+        idx = centros.index(cercano)
+
+        if idx < len(letras):
+            texto += letras[idx]
+
+    return texto
+
+
+with gr.Blocks() as demo:
+
+    gr.Markdown("# Decodificador de tonos estilo radio digital")
+
+    audio = gr.Audio(type="filepath")
+
+    boton = gr.Button("Decodificar")
+
+    salida = gr.Textbox(lines=10)
+
+    boton.click(
+        decodificar,
+        inputs=audio,
+        outputs=salida
+    )
+
+demo.launch()