In this paper, we investigate multi-speaker emotional acoustic modeling methods for convolutional neural network (CNN) based speech synthesis system. For emotion modeling, we extend to the speech synthesis system that learns a latent embedding space of emotion, derived from a desired emotional identity, and we use emotion code and mel-frequency spectrogram as an emotion identity. In order to model speaker variation in a text-to-speech (TTS) system, we use speaker representations such as trainable speaker embedding and speaker code. We have implemented speech synthesis systems combining speaker representation and emotion representation and compared them by experiments. Experimental results have demonstrated that the multi-speaker emotional speech synthesis approach using trainable speaker embedding and emotion representation from mel spectrogram achieves higher performance when compared with other approaches in terms of naturalness, speaker similarity, and emotion similarity.

A novel neural network architecture, called two-stage network (TSN), with a multi-objective learning (MOL) method for an efficient boosting strategy (BS) is proposed for speech enhancement. BS is an ensemble method using multiple base predictions (MBPs) for better final prediction. Because of the necessity for MBPs, the computational cost and model size of BS-based methods are greater than those of a single model. In overcoming this, TSN first obtains MBPs from a single deep neural network. Then, to obtain better final prediction, the convolution layers of TSN aggregate not only MBPs but also some auxiliary information such as contextual information, while adaptively filtering out some unnecessary information, e.g., poor base predictions. At the training phase, the MOL enables all stages of TSN to learn jointly, whereas allowing the TSN framework to embed a BS. Our experimental results confirm that the embedded BS leads TSN to outperform other baseline methods with a reasonably low computational cost and model size.

Voice activity detection (VAD) classifies incoming signal segments into speech or background noise; its performance is crucial in various speech-related applications. Although speech-signal context is a relevant VAD asset, its usefulness varies in unpredictable noise environments. Therefore, its usage should be adaptively adjustable to the noise type. This letter improves the use of context information by using an adaptive context attention model (ACAM) with a novel training strategy for effective attention, which weights the most crucial parts of the context for proper classification. Experiments in real-world scenarios demonstrate that the proposed ACAM-based VAD outperforms the other baseline VAD methods.

In this paper, we propose a new probabilistic approach to pulse transit time (PTT) estimation using a Gaussian distribution model. It is motivated basically by the hypothesis that PTTs normalized by RR intervals follow the Gaussian distribution. To verify the hypothesis, we demonstrate the effects of arterial compliance on the normalized PTTs using the Moens-Korteweg equation. Furthermore, we observe a Gaussian distribution of the normalized PTTs on real data. In order to estimate the PTT using the hypothesis, we first assumed that R-waves in the electrocardiogram (ECG) can be correctly identified. The R-waves limit searching ranges to detect pulse peaks in the photoplethysmogram (PPG) and to synchronize the results with cardiac beats--i.e., the peaks of the PPG are extracted within the corresponding RR interval of the ECG as pulse peak candidates. Their probabilities of being the actual pulse peak are then calculated using a Gaussian probability function. The parameters of the Gaussian function are automatically updated when a new pulse peak is identified. This update makes the probability function adaptive to variations of cardiac cycles. Finally, the pulse peak is identified as the candidate with the highest probability. The proposed approach is tested on a database where ECG and PPG waveforms are collected simultaneously during the submaximal bicycle ergometer exercise test. The results are promising, suggesting that the method provides a simple but more accurate PTT estimation in real applications.

In this paper, we propose a novel method for enhancing pulse contour analysis-based arterial stiffness estimation using a simple and low-complexity photoplethysmographic parameter (P2Ocd). The method first eliminates baseline wanders in the digital volume pulse (DVP) by applying a simple morphological filter. The filtered DVP signal is then transformed into a slope sum function signal to simplify the pulse peak detection process by enhancing the upslope of the DVP signal while suppressing its downslope. An adaptive thresholding scheme is applied to detect pulse peaks from the transformed signal. Pulse onsets are then identified as the minimum values between consecutive pulse peaks. The P2Ocd is finally calculated by dividing the time interval between the pulse peak and the pulse onset by the pulse length. In order to assess the agreement of the P2Ocd with an established technique, brachial-ankle pulse wave velocity, we performed Bland-Altman and correlation analyses. Furthermore, we evaluated the P2Ocd-based arterial stiffness estimation in terms of prediction accuracy (% error rate) and repeatability (coefficient of variation). The results show that the proposed measurement agrees well with the established technique and shows a high repeatability; it also has a better predictive accuracy than that of conventional methods. In addition, we show that the proposed parameter further improves the predictive accuracy by combining it with age. The proposed method is therefore highly applicable to small ubiquitous healthcare applications.

This letter proposes a bandwidth extension (BWE) method using the cepstral envelope coding and duplication of quantized wideband (WB) signals by means of analysis -by-synthesis (AbS) for super-wideband (SWB) coders. In the proposed method, a high frequency band is generated by utilizing the quantized cepstral coefficients extracted from the envelope and the quantized modified discrete cosine transform (MDCT) shape of the wideband signal. The proposed method is compared with the latest G.718 SWB codec and the experimental results show that the proposed method outperforms the baseline codec both in subjective listening tests and objective performance measures.

Interactive audio services (IASs) usually provide users with audio editing functionality and they can render their own sounds according to their preference. For IASs, the spatial audio object coding (SAOC) is an appropriate multichannel coding tool that satisfies most of the required functionalities with relatively low bit rate. Nevertheless, the SAOC usually fails to remove a specific object successfully, especially the vocal object in the case of the Karaoke service. In addition, to expand the service to mobile environments, lower bit rate and complexity are required. Thus, we propose a new SAOC vocal harmonic coding technique to improve the background music quality in the Karaoke service. Namely, utilizing the harmonic information of the vocal object, we removed the harmonics of the vocal object remaining in the background music. Our experimental results confirm that the background music quality is improved by the proposed algorithm even with the low bit rate and complexity.

In this paper, a new algorithm for encoding spectral envelope in G.729.1 for VoIP is proposed. In the TDAC part of G.729.1, the spectral envelope and MDCT coefficients extracted in the weighted CELP coding error (lower-band) and the higher-band input signal are encoded. In order to reduce allocation bits for spectral envelope coding, a new algorithm using sub-band correlation between adjacent frames is proposed. In addition, a new bit allocation strategy using reduced bits from the proposed algorithm is also proposed. The performance of the proposed algorithm is evaluated by sound quality and bit reduction rates in clean and frame loss conditions.

An interactive audio service is a new conceptual audio service that provides the users with opportunities for a variety of experiences on the alternative and advanced audio services. In the interactive audio service, users can freely control various audio objects to make their own audio sounds. A spatial audio object coding (SAOC) is a useful technology that can support most parts of the interactive audio service with a relatively low bit-rate, but is very poor to perfect gain control of a certain audio object, i.e., the target audio object. In this paper, the SAOC with a two-step coding structure is proposed to efficiently handle the target audio object as well as the normal audio objects. A transform coded excitation (TCX) based residual coding scheme is presented in the context of the sound quality enhancement. From experimental results, it can be noted that the various audio objects can be successfully handled with respect to the bit-rate and the sound quality by using the proposed two-step coding structure SAOC.

This paper propose adaptive noisy signal reduction method on smart devices for voice communication service in pervasive home environment. In order to reduce noisy signals efficiently on smart devices we provide Wiener filtering technique based on input-SNR estimation method with preprocessing on smart home devices.

The primary aim of this paper is to present new extracting scheme of specific scenes considering heterogeneous terminals, which have different screen sizes for providing viewer centric personal broadcasting IP-TV service.

This paper describes a new robot photographer system which can interact with people. The goal of this research is to make the system act like a human photographer. This system is based on a mobile robot having capabilities of wireless communication and stereo vision. It recognizes waving hands of people, moves toward them, and takes pictures with designated compositions. The pictures are transmitted to personal computers over the wireless network. In comparison with previous researches, the most unique things of this system are human interaction and user preference. To realize these properties, this new robot photographer system is considered optical property of lens and applied interesting vision algorithm which was never tried before for previous robot photographer systems

This paper describes the prosodic features of Korean dialogue speech. With 25 sentences for scheduling, one speaker uttered in two manners, viz. 'read' and 'dialogue'. The main discriminating features would be some aspects in speech rate and boundary signal. We discriminated each prosodic phrase in a sentence to investigate pre-boundary, boundary, and post-boundary features. drastic characteristics than that in read. We can see that the boundary syllables of dialogue seem to be 2.3 times longer than that in preboundary syllable. The final syllables are about 1.7 times longer than prefinal syllables. Pitch

A new, simple speech detection algorithm is implemented and applied to isolated Korean utterances with encouraging results. The algorithm makes use of decision parameters and threshold values based on three features: the logarithmic energy, the zero crossing rate, and the modified zero crossing rate. All threshold values are fixed through the training procedure by applying a discrete optimization technique for the prepared training data set. The vocabulary for the experiment includes Korean digits and some other control commands designed for an automatic dialing system. Tested on 384 utterances from three male and three female speakers, the algorithm produces 4.6 ms average error while the well-known Rabiner and Sambur's endpoint detection algorithm gives 18.0 ms average error for the same speaker-independent data set. It is also shown that the new algorithm can be improved by training for male and female separately.>