基于PSO优化GRNN的语音转换方法

提出了一种基于粒子群算法PSO优化广义回归神经网络GRNN模型的语音转换方法。首先,该方法利用训练语音的声道和激励源的个性化特征参数分别训练两个GRNN,得到GRNN的结构参数;然后,利用PSO对GRNN的结构参数进行优化,减少人为因素对转换结果的影响;最后,对语音的韵律特征、基音轮廓和能量分别进行了线性转换,使得转换后的语音包含更多源语音的个性化特征信息。主客观实验结果表明:与径向基神经网络RBF和GRNN相比,使用本文提出的转换模型获得的转换语音的自然度和似然度都得到了很大的提升,谱失真率明显降低并且更接近于目标语音。     

A pitch synchronous approach to design voice conversion system using source-filter correlation

We propose a pitch synchronous approach to design the voice conversion system taking into account the correlation between the excitation signal and vocal tract system characteristics of speech production mechanism. The glottal closure instants (GCIs) also known as epochs are used as anchor points for analysis and synthesis of the speech signal. The Gaussian mixture model (GMM) is considered to be the state-of-art method for vocal tract modification in a voice conversion framework. However, the GMM based models generate overly-smooth utterances and need to be tuned according to the amount of available training data. In this paper, we propose the support vector machine multi-regressor (M-SVR) based model that requires less tuning parameters to capture a mapping function between the vocal tract characteristics of the source and the target speaker. The prosodic features are modified using epoch based method and compared with the baseline pitch synchronous overlap and add (PSOLA) based method for pitch and time scale modification. The linear prediction residual (LP residual) signal corresponding to each frame of the converted vocal tract transfer function is selected from the target residual codebook using a modified cost function. The cost function is calculated based on mapped vocal tract transfer function and its dynamics along with minimum residual phase, pitch period and energy differences with the codebook entries. The LP residual signal corresponding to the target speaker is generated by concatenating the selected frame and its previous frame so as to retain the maximum information around the GCIs. The proposed system is also tested using GMM based model for vocal tract modification. The average mean opinion score (MOS) and ABX test results are 3.95 and 85 for GMM based system and 3.98 and 86 for the M-SVR based system respectively. The subjective and objective evaluation results suggest that the proposed M-SVR based model for vocal tract modification combined with modified residual selection and epoch based model for prosody modification can provide a good quality synthesized target output. The results also suggest that the proposed integrated system performs slightly better than the GMM based baseline system designed using either epoch based or PSOLA based model for prosody modification.     

Comparing ANN and GMM in a voice conversion framework

In this paper, we present a comparative analysis of artificial neural networks (ANNs) and Gaussian mixture models (GMMs) for design of voice conversion system using line spectral frequencies (LSFs) as feature vectors. Both the ANN and GMM based models are explored to capture nonlinear mapping functions for modifying the vocal tract characteristics of a source speaker according to a desired target speaker. The LSFs are used to represent the vocal tract transfer function of a particular speaker. Mapping of the intonation patterns (pitch contour) is carried out using a codebook based model at segmental level. The energy profile of the signal is modified using a fixed scaling factor defined between the source and target speakers at the segmental level. Two different methods for residual modification such as residual copying and residual selection methods are used to generate the target residual signal. The performance of ANN and GMM based voice conversion (VC) system are conducted using subjective and objective measures. The results indicate that the proposed ANN-based model using LSFs feature set may be used as an alternative to state-of-the-art GMM-based models used to design a voice conversion system.     

Robust processing techniques for voice conversion

Differences in speaker characteristics, recording conditions, and signal processing algorithms affect output quality in voice conversion systems. This study focuses on formulating robust techniques for a codebook mapping based voice conversion algorithm. Three different methods are used to improve voice conversion performance: confidence measures, pre-emphasis, and spectral equalization. Analysis is performed for each method and the implementation details are discussed. The first method employs confidence measures in the training stage to eliminate problematic pairs of source and target speech units that might result from possible misalignments, speaking style differences or pronunciation variations. Four confidence measures are developed based on the spectral distance, fundamental frequency (f₀) distance, energy distance, and duration distance between the source and target speech units. The second method focuses on the importance of pre-emphasis in line-spectral frequency (LSF) based vocal tract modeling and transformation. The last method, spectral equalization, is aimed at reducing the differences in the source and target long-term spectra when the source and target recording conditions are significantly different. The voice conversion algorithm that employs the proposed techniques is compared with the baseline voice conversion algorithm with objective tests as well as three subjective listening tests. First, similarity to the target voice is evaluated in a subjective listening test and it is shown that the proposed algorithm improves similarity to the target voice by 23.0%. An ABX test is performed and the proposed algorithm is preferred over the baseline algorithm by 76.4%. In the third test, the two algorithms are compared in terms of the subjective quality of the voice conversion output. The proposed algorithm improves the subjective output quality by 46.8% in terms of mean opinion score (MOS).     

采用深度信念网络的语音转换方法

对说话人语音个性特征信息的表征和提取进行了深入研究,提出了一种基于深度信念网络（Deep Belief Nets,DBN）的语音转换方法。分别用提取出的源说话人和目标说话人语音频谱参数来训练DBN,分别得到其在高阶空间的语音个性特征表征;通过人工神经网络（Artificial Neural Networks,ANN）来连接这两个高阶空间并进行特征转换;使用基于目标说话人数据训练出的DBN来对转换后的特征信息进行逆处理得到转换后语音频谱参数,合成转换语音。实验结果表明,与传统的基于GMM方法相比,该方法效果更好,转换语音音质和相似度同目标语音更接近。     

一种基于正弦激励的线性预测模型的语音转换方法

在正弦激励模型的线性预测(LP)残差转换的基础上,提出了一种改进语音特征转换性能的语音转换方法.基于线性预测分析和综合的构架,该方法一方面通过谱包络估计声码器提取源说话人的线性预测编码(LPC)倒谱包络,并使用双线性变换函数实现倒谱包络的转换;另一方面由谐波正弦模型对线性预测残差信号建模和分解,采用基音频率变换将源说话人的残差信号转换为近似目标说话人的残差信号.最后由修正后的残差信号激励时变滤波器得到转换语音,滤波器参数通过转换得到的LPC倒谱包络实时更新.实验结果表明,该方法在主观和客观测试中都具有良好的结果,能有效地转换说话人声音特征,获得高相似度的转换语音.     

Improvement of phone recognition accuracy using speech mode classification

In this work, we have developed a speech mode classification model for improving the performance of phone recognition system (PRS). In this paper, we have explored vocal tract system, excitation source and prosodic features for development of speech mode classification (SMC) model. These features are extracted from voiced regions of a speech signal. In this study, conversation, extempore, and read speech are considered as three different modes of speech. The vocal tract component of speech is extracted using Mel-frequency cepstral coefficients (MFCCs). The excitation source features are captured through Mel power differences of spectrum in sub-bands (MPDSS) and residual Mel-frequency cepstral coefficients (RMFCCs) of the speech signal. The prosody information is extracted from pitch and intensity. Speech mode classification models are developed using above described features independently, and in fusion. The experiments carried out on Bengali speech corpus to analyze the accuracy of the speech mode classification model using the artificial neural network (ANN), naive Bayes, support vector machines (SVMs) and k-nearest neighbor (KNN). We proposed four classification models which are combined using maximum voting approach for optimal performance. From the results, it is observed that speech mode classification model developed using the fusion of vocal tract system, excitation source and prosodic features of speech, yields the best performance of 98%. Finally, the proposed speech mode classifier is integrated to the PRS, and the accuracy of phone recognition system is observed to be improved by 11.08%.     

基于话者无关模型的说话人转换方法

提出一种基于话者无关模型的说话人转换方法.考虑到音素信息共同存在于所有说话人的语音中,假设存在一个可以用高斯混合模型来描述的话者无关空间,且可用分段线性变换来描述该空间到各说话人相关空间之间的映射关系.在一个多说话人的数据库上,用话者自适应训练算法来训练模型,并在转换阶段使用源目标说话人空间到话者无关空间的变换关系来构造源与目标之间的特征变换关系,快速、灵活的构造说话人转换系统.通过主观测听实验来验证该算法相对于传统的基于话者相关模型方法的优点.     

High quality voice conversion using prosodic and high-resolution spectral features

Voice conversion methods have advanced rapidly over the last decade. Studies have shown that speaker characteristics are captured by spectral feature as well as various prosodic features. Most existing conversion methods focus on the spectral feature as it directly represents the timbre characteristics, while some conversion methods have focused only on the prosodic feature represented by the fundamental frequency. In this paper, a comprehensive framework using deep neural networks to convert both timbre and prosodic features is proposed. The timbre feature is represented by a high-resolution spectral feature. The prosodic features include F0, intensity and duration. It is well known that DNN is useful as a tool to model high-dimensional features. In this work, we show that DNN initialized by our proposed autoencoder pretraining yields good quality DNN conversion models. This pretraining is tailor-made for voice conversion and leverages on autoencoder to capture the generic spectral shape of source speech. Additionally, our framework uses segmental DNN models to capture the evolution of the prosodic features over time. To reconstruct the converted speech, the spectral feature produced by the DNN model is combined with the three prosodic features produced by the DNN segmental models. Our experimental results show that the application of both prosodic and high-resolution spectral features leads to quality converted speech as measured by objective evaluation and subjective listening tests.     

Text-independent speaker identification based on selection of the most similar feature vectors

The speaker recognition has been one of the interesting issues in signal and speech processing over the last few decades. Feature selection is one of the main parts of speaker recognition system which can improve the performance of the system. In this paper, we have proposed two methods to find MFCCs feature vectors with the highest similar that is applied to text independent speaker identification system. These feature vectors show individual properties of each person’s vocal tract that are mostly repeated. They are used to build speaker’s model and to specify decision boundary. We applied MFCC of each window over main signal as a feature vector and used clustering to obtain feature vectors with the highest similar. The Speaker identification experiments are performed using the ELSDSR database that consists of 22 speakers (12 male and 10 female) and Neural Network is used as a classifier. The effect of three main parameters have been considered in two proposed methods. Experimental results indicate that the performance of speaker identification system has been improved in accuracy and time consumption term.     

改进的跨语种语音合成模型自适应方法

统计参数语音合成中的跨语种模型自适应主要应用于目标说话人语种与源模型语种不同时,使用目标发音人少量语音数据快速构建具有其音色特征的源模型语种合成系统。本文对传统的基于音素映射和三音素模型的跨语种自适应方法进行改进,一方面通过结合数据挑选的音素映射方法以提高音素映射的可靠性,另一方面引入跨语种的韵律信息映射以弥补原有方法中三音素模型在韵律表征上的不足。在中英文跨语种模型自适应系统上的实验结果表明,改进后系统合成语音的自然度与相似度相对传统方法都有了明显提升。     

Parametric Formant Modelling and Transformation in Voice Conversion

This paper presents a method for the estimation and mapping of parametric models of speech resonance at formants for voice conversion. The spectral features at formants that contribute to voice characteristics are the trajectories of the frequencies, the bandwidths and intensities of the resonance at formants. The formant features are extracted from the poles of a linear prediction (LP) model of speech. The statistical distributions of formants are modelled by a two-dimensional hidden Markov model (HMM) spanning the time and frequency dimensions. Experimental results are presented which show a close match between HMM-based formant models and the histograms of formants. For voice conversion two alternative methods are explored for mapping the formants of a source speaker to those of a target speaker. The first method is based on an adaptive formant-tracking warping of the frequency response of the LP model and the second method is based on the rotation of the poles of the LP model of speech. Both methods transform all spectral parameters of the resonance at formants of the source speaker towards those of the target speaker. In addition, the issues affecting the selection of the warping ratios for the mapping functions are investigated. Experimental results of formant estimation and perceptual evaluation of voice morphing based on parametric formant models are presented.     

采用STRAIGHT模型和深度信念网络的语音转换方法

提出一种将STRAIGHT模型和深度信念网络DBN相结合实现语音转换的方式。首先,通过STRAIGHT模型提取出源说话人和目标说话人的语音频谱参数,用提取的频谱参数分别训练两个DBN得到语音高阶空间的个性特征信息;然后,用人工神经网络ANN将两个具有高阶特征的空间连接并进行特征转换;最后,用基于目标说话人数据训练出的DBN来对转换后的特征信息进行逆处理得到语音频谱参数,并用STRAIGHT模型合成具有目标说话人个性化特征的语音。实验结果表明,采用此种方式获得的语音转换效果要比传统的采用GMM实现语音转换更好,转换后的语音音质和相似度与目标语音更接近。     

Speaker verification using excitation source information

In this work we develop a speaker recognition system based on the excitation source information and demonstrate its significance by comparing with the vocal tract information based system. The speaker-specific excitation information is extracted by the subsegmental, segmental and suprasegmental processing of the LP residual. The speaker-specific information from each level is modeled independently using Gaussian mixture modeling—universal background model (GMM-UBM) modeling and then combined at the score level. The significance of the proposed speaker recognition system is demonstrated by conducting speaker verification experiments on the NIST-03 database. Two different tests, namely, Clean test and Noisy test are conducted. In case of Clean test, the test speech signal is used as it is for verification. In case of Noisy test, the test speech is corrupted by factory noise (9 dB) and then used for verification. Even though for Clean test case, the proposed source based speaker recognition system still provides relatively poor performance than the vocal tract information, its performance is better for Noisy test case. Finally, for both clean and noisy cases, by providing different and robust speaker-specific evidences, the proposed system helps the vocal tract system to further improve the overall performance.     

基于翻转梅尔频率倒谱系数的语音变调检测方法

语音变调常用于掩盖说话人身份,各种变声软件的出现使得说话人身份伪装变得更加容易。针对现有变调语音检测方法无法判断语音是经过了何种变调操作（升调或降调）的问题,通过分析语音变调在信号频谱,尤其是高频区域留下的痕迹,提出了基于翻转梅尔倒谱系数（IMFCC）统计矩特征的电子变调语音检测方法。首先,提取各语音帧IMFCC及其一阶差分;然后,计算其统计均值;最后,在该统计特征上利用支持向量机（SVM）多分类器的设计来区分原始语音、升调语音和降调语音。在TIMIT和NIST语音集上的实验结果表明,所提方法无论对于原始语音、升调语音还是降调语音都具有良好的检测性能。与MFCC作为特征构造的基线系统相比,所设计的特征的方法明显提高了变调操作的识别率。在较少的训练资源的情况下,所提方法也获得了比基于卷积神经网络（CNN）的框架更好的性能;此外,在不同数据集和不同变调方法上也都取得了较好的泛化性能。     

Voice conversion using duration-embedded bi-HMMs for expressive speech synthesis

This paper presents an expressive voice conversion model (DeBi-HMM) as the post processing of a text-to-speech (TTS) system for expressive speech synthesis. DeBi-HMM is named for its duration-embedded characteristic of the two HMMs for modeling the source and target speech signals, respectively. Joint estimation of source and target HMMs is exploited for spectrum conversion from neutral to expressive speech. Gamma distribution is embedded as the duration model for each state in source and target HMMs. The expressive style-dependent decision trees achieve prosodic conversion. The STRAIGHT algorithm is adopted for the analysis and synthesis process. A set of small-sized speech databases for each expressive style is designed and collected to train the DeBi-HMM voice conversion models. Several experiments with statistical hypothesis testing are conducted to evaluate the quality of synthetic speech as perceived by human subjects. Compared with previous voice conversion methods, the proposed method exhibits encouraging potential in expressive speech synthesis.     

基于卷积非负矩阵分解的语音转换方法

为了在语音转换过程中充分考虑语音的帧间相关性,提出了一种基于卷积非负矩阵分解的语音转换方法.卷积非负矩阵分解得到的时频基可较好地保存语音信号中的个人特征信息及帧间相关性.利用这一特性,在训练阶段,通过卷积非负矩阵分解从训练数据中提取源说话人和目标说话人相匹配的时频基.在转换阶段,通过时频基替换实现对源说话人语音的转换.相对于传统方法,本方法能够更好地保存和转换语音帧间相关性.实验仿真及主、客观评价结果表明,与基于高斯混合模型、状态空间模型的语音转换方法相比,该方法具有更好的转换语音质量和转换相似度.     

Performance of new voice conversion systems based on GMM models and applied to Arabic language

Voice conversion (VC) consists in modifying the source speaker’s voice toward the voice of the target speaker. In our paper, we are interested in calculating the performance of a conversion system based on GMM, applied to the Arabic language, by exploiting both the information of the pitch dynamics and the spectrum. We study three approaches to obtain the global conversion function of the pitch and the overall spectrum, using the joint probability model. In the first approach, we calculate the joint conversion of pitch and spectrum. In the second approach, the pitch is calculated by linear conversion. In the third approach, we use the relationship between the pitch and the spectrum. For the conversion of noise we use a new technique that consists in modeling the noise of the voiced or unvoiced frames by GMMs. We use the HNM for analysis/synthesis and a regularized discrete cepstrum in order to estimate the spectrum of the speech signal.     

A novel prosodic-information synthesizer based on recurrent fuzzy neural network for the Chinese TTS system.

In this paper, a new technique for the Chinese text-to-speech (TTS) system is proposed. Our major effort focuses on the prosodic information generation. New methodologies for constructing fuzzy rules in a prosodic model simulating human's pronouncing rules are developed. The proposed Recurrent Fuzzy Neural Network (RFNN) is a multilayer recurrent neural network (RNN) which integrates a Self-cOnstructing Neural Fuzzy Inference Network (SONFIN) into a recurrent connectionist structure. The RFNN can be functionally divided into two parts. The first part adopts the SONFIN as a prosodic model to explore the relationship between high-level linguistic features and prosodic information based on fuzzy inference rules. As compared to conventional neural networks, the SONFIN can always construct itself with an economic network size in high learning speed. The second part employs a five-layer network to generate all prosodic parameters by directly using the prosodic fuzzy rules inferred from the first part as well as other important features of syllables. The TTS system combined with the proposed method can behave not only sandhi rules but also the other prosodic phenomena existing in the traditional TTS systems. Moreover, the proposed scheme can even find out some new rules about prosodic phrase structure. The performance of the proposed RFNN-based prosodic model is verified by imbedding it into a Chinese TTS system with a Chinese monosyllable database based on the time-domain pitch synchronous overlap add (TD-PSOLA) method. Our experimental results show that the proposed RFNN can generate proper prosodic parameters including pitch means, pitch shapes, maximum energy levels, syllable duration, and pause duration. Some synthetic sounds are online available for demonstration.     

On robustness of speech based biometric systems against voice conversion attack

Voice conversion (VC) approach, which morphs the voice of a source speaker to be perceived as spoken by a specified target speaker, can be intentionally used to deceive the speaker identification (SID) and speaker verification (SV) systems that use speech biometric. Voice conversion spoofing attacks to imitate a particular speaker pose potential threat to these kinds of systems. In this paper, we first present an experimental study to evaluate the robustness of such systems against voice conversion disguise. We use Gaussian mixture model (GMM) based SID systems, GMM with universal background model (GMM-UBM) based SV systems and GMM supervector with support vector machine (GMM-SVM) based SV systems for this. Voice conversion is conducted by using three different techniques: GMM based VC technique, weighted frequency warping (WFW) based conversion method and its variation, where energy correction is disabled (WFW⁻). Evaluation is done by using intra-gender and cross-gender voice conversions between fifty male and fifty female speakers taken from TIMIT database. The result is indicated by degradation in the percentage of correct identification (POC) score in SID systems and degradation in equal error rate (EER) in all SV systems. Experimental results show that the GMM-SVM SV systems are more resilient against voice conversion spoofing attacks than GMM-UBM SV systems and all SID and SV systems are most vulnerable towards GMM based conversion than WFW and WFW⁻ based conversion. From the results, it can also be said that, in general terms, all SID and SV systems are slightly more robust to voices converted through cross-gender conversion than intra-gender conversion. This work extended the study to find out the relationship between VC objective score and SV system performance in CMU ARCTIC database, which is a parallel corpus. The results of this experiment show an approach on quantifying objective score of voice conversion that can be related to the ability to spoof an SV system.