Add python example: Vosk + CMU Pronunciation Feedback for Children

python/example/phoneme_feedback_kiddo.py

@@ -0,0 +1,87 @@
+# phoneme_feedback_kiddo.py
+# Educational Example – Vosk + CMU Pronouncing Dictionary for Pronunciation Feedback
+# Author: Aparna V Sunil
+
+import os
+import wave
+import json
+import nltk
+from vosk import Model, KaldiRecognizer
+from nltk.corpus import cmudict
+
+# Ensure CMU dictionary is downloaded
+nltk.download('cmudict')
+pron_dict = cmudict.dict()
+
+# Setup Vosk model path
+MODEL_PATH = "model"  # Replace with your actual model folder name (e.g., vosk-model-en-us-0.22)
+AUDIO_FILE = "test.wav"  # Replace with a .wav file of a child saying the target word
+
+# Target word expected from child
+TARGET_WORD = "elephant"
+
+# Load expected phonemes
+expected_phonemes = pron_dict.get(TARGET_WORD.lower(), [[]])[0]
+expected_clean = [p.strip("0123456789").lower() for p in expected_phonemes]
+
+print(f"\n🎯 Target Word: {TARGET_WORD}")
+print(f"✅ Expected Phonemes: {expected_clean}")
+
+# Initialize Vosk recognizer
+if not os.path.exists(MODEL_PATH):
+    print(f"❌ Vosk model not found at: {MODEL_PATH}")
+    exit(1)
+
+model = Model(MODEL_PATH)
+wf = wave.open(AUDIO_FILE, "rb")
+
+if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getframerate() != 16000:
+    print("❌ Audio file must be WAV 16KHz Mono PCM.")
+    exit(1)
+
+recognizer = KaldiRecognizer(model, wf.getframerate())
+
+# Run speech recognition
+recognized_text = ""
+while True:
+    data = wf.readframes(4000)
+    if len(data) == 0:
+        break
+    if recognizer.AcceptWaveform(data):
+        result = json.loads(recognizer.Result())
+        recognized_text += result.get("text", "") + " "
+
+recognized_text = recognized_text.strip().lower()
+print(f"\n🎙️ Recognized Speech: {recognized_text}")
+
+# Extract recognized phonemes
+recognized_phonemes = []
+for word in recognized_text.split():
+    phones = pron_dict.get(word.lower())
+    if phones:
+        recognized_phonemes.extend(phones[0])
+
+recognized_clean = [p.strip("0123456789").lower() for p in recognized_phonemes]
+
+print(f"🗣️ Recognized Phonemes: {recognized_clean}")
+
+# Compare phonemes
+print("\n📌 Phoneme Feedback:")
+matched = set(expected_clean).intersection(set(recognized_clean))
+missing = list(set(expected_clean) - set(recognized_clean))
+
+for phoneme in expected_clean:
+    if phoneme in matched:
+        print(f"✅ {phoneme}")
+    else:
+        print(f"❌ Missing or incorrect: {phoneme}")
+
+similarity = len(matched) / len(expected_clean) if expected_clean else 0
+print(f"\n🔎 Similarity Score: {similarity:.2f}")
+
+if similarity == 1.0:
+    print("🎉 Perfect pronunciation!")
+elif similarity >= 0.6:
+    print("👍 Good attempt, but some phonemes need work.")
+else:
+    print("❗ Needs improvement. Try again!")