I don't think so. Kana just don't have enough entropy compares to Kanji. A kanji can be compose with up to twenty strokes with high variety of stroke patterns. Those excessive complexity make it identifiable even in extreme situations. (Blurry or tainted or whatever situation). In some case, a kanji with half of its size masked still be decoded without any ambiguity. But this will never work with an voice based language.