Addable Stress Speech Recognition with Multiplexing HMM: Training and Non-training Decision