This paper presents a deep neural network (DNN)-based unit selection method for waveform concatenation speech synthesis using frame-sized speech segments. In this method, three DNNs are adopted to calculate target costs and concatenation costs respectively for selecting frame-sized candidate units. The first DNN is built in the same way as the DNN-based statistical parametric speech synthesis, which predicts target acoustic features given linguistic context inputs.
The distance between the acoustic features of a candidate unit and the predicted ones for a target unit is calculated as the target cost. The other two DNNs are constructed to predict the acoustic features at current frame using its context features and the acoustic features of previous frames. At synthesis time, these two DNNs are employed to calculate the concatenation cost for each candidate unit given its preceding units. Experimental results show that our proposed method can achieve better naturalness than the hidden Markov model (HMM)-based frame selection method and the HMM-based parametric synthesis method.