For producing high quality synthesis, a concatenation-based Text-to-Speech (TTS) system usually requires a large number of segmental units to cover various acoustic-phonetic contexts. However, careful manual labeling and segmentation by human experts, which is still the most reliable way to prepare such units, is labor intensive. In this paper we adopt a two-step procedure to automate the labeling, segmentation and refinement process. In the first step, coarse segmentation of speech data is performed by aligning speech signals with the corresponding sequence of Hidden Markov Models (HMMs). Then in the second step, segment boundaries are refined with a proposed Context-Dependent Boundary Model (CDBM). Classification and Regression Tree (CART) is adopted to organize available data into a structured hierarchical tree, where acoustically similar boundaries are clustered together to train tied CDBM models for boundary refinement. Optimal CDBM parameters and training conditions are found through a series of experimental studies. Comparing with manual segmentation reference, segmentation accuracy (within a tolerance of 20 ms) is improved by the CDBMs from 78.1% (baseline) to 94.8% in Mandarin Chinese and from 81.4% to 92.7% in English, with about 1,000 manually segmented sentences used in training the models. To further reduce the amount of manual data for training CDBMs of a new speaker, we adapt a well-trained CDBM via efficient adaptation algorithms. With only 10-20 manually segmented sentences as adaptation data, the adapted CDBM achieves a segmentation accuracy of 90%.

This paper proposes a SNR estimation scheme based on Gibbs sampler. This scheme can estimate SNR using a very short received sequence, and does not require any prior information of the transmitted symbol. Compared with the existing estimators, the performance of this method is better when real SNR is larger than 5 dB in both single path channel and multi-path channel.

In this paper, we present a new transmit diversity scheme with power control by using two transmit antennas in which the power control unit is added to adaptively suit the channel fading variation. Compared to the transmit diversity scheme (STD, one space time coding scheme) proposed by Alamouti and the traditional maximal ratio combining (MRC) diversity scheme employed at the receiver, simulation results indicate that the new scheme has considerable performance gain. We also discuss the effects of the imperfect channel parameter estimation on the performance of the system. Simulation results show that the new system is more robust to the estimation error of channel fading parameters than the STD. As the signal to noise ratio is relatively high, the diversity scheme with power control is more sensitive to the channel estimation error compared to the MRC. But when the channel estimation accuracy is relatively high, the diversity scheme with power control still has better performance than the ideal MRC as the BER is about 1 10-3.

In this paper, we consider the subchannel detection problem in forward links for the multicarrier DS-CDMA system when some different subchannel allocation policies are used. An optimal subchannel decision algorithm is proposed based on the maximum-likelihood (ML) criterion. Theoretical analysis and simulation results are presented. Furthermore, we discuss the parameter selection problem on the length of the training sequences in the subchannel allocation scheme in [8],[12] by using the proposed ML detection algorithm. The results show that the subchannel allocation scheme in [8],[12] is feasible since only a few symbols overhead is required.

A leakage-aware Coordinated Scheduling/Coordinated Beamforming (CS/CB) scheme for heterogeneous networks with layered limited feedback is proposed. In particular, all pico cells cooperatively select an optimal beamforming vector for the macro cell within a CoMP cluster so as to minimizing leakage power from the macro cell. Simulations show that the proposed scheme outperforms the conventional non-CoMP scheme with perfect channel state information at teansmitter (CSIT). Furthermore, the feedback amount and scheduler complexity is decreased greatly.

In this letter, we first analyze the delay of the WLAN discovery scheme specified for 3GPP and WLAN interworking networks. Theoretical analysis indicates that the delay of the discovery scheme given by 3GPP increases linearly with the number of WLAN channels needed to be scanned. To reduce the discovery delay, we then propose an effective WLAN discovery scheme utilizing the cellular network to aid in the broadcasting the information of WLANs. Thus the number of WLAN channels required to be scanned for users is greatly reduced. The effectiveness of the proposed scheme is demonstrated by analysis and simulation.