Vocal Tract Filter Research Articles

Alternating minimisation for glottal inverse filtering

A new method is proposed for solving the glottal inverse filtering (GIF) problem. The goal of GIF is to separate an acoustical speech signal into two parts: the glottal airflow excitation and the vocal tract filter. To recover such information one has to deal with a blind deconvolution problem. This ill-posed inverse problem is solved under a deterministic setting, considering unknowns on both sides of the underlying operator equation. A stable reconstruction is obtained using a double regularization strategy, alternating between fixing either the glottal source signal or the vocal tract filter. This enables not only splitting the nonlinear and nonconvex problem into two linear and convex problems, but also allows the use of the best parameters and constraints to recover each variable at a time. This new technique, called alternating minimization glottal inverse filtering (AM-GIF), is compared with two other approaches: Markov chain Monte Carlo glottal inverse filtering (MCMC-GIF), and iterative adaptive inverse filtering (IAIF), using synthetic speech signals. The recent MCMC-GIF has good reconstruction quality but high computational cost. The state-of-the-art IAIF method is computationally fast but its accuracy deteriorates, particularly for speech signals of high fundamental frequency (F0). The results show the competitive performance of the new method: With high F0, the reconstruction quality is better than that of IAIF and close to MCMC-GIF while reducing the computational complexity by two orders of magnitude.

Modeling and joint estimation of glottal source and vocal tract filter by state-space methods

Abstract Accurate estimation of the glottal source from a voiced sound is a difficult blind separation problem in speech signal processing. In this work, state-space methods are investigated to enhance the joint estimation of the glottal source and the vocal tract information. The aim of this paper is twofold. First, a stochastic glottal source is proposed, based on deterministic Liljencrants–Fant model and ruled by a stochastic difference equation. Such a representation allows to accurately capture any perturbation occurring at glottal level in real voices. A state-space voice model is formulated considering the stochastic glottal source. Then, combining this voice model and the state-space framework, an inverse filtering method is developed that allows to jointly estimate both glottal source and vocal tract filter. The performance of this method is studied by means of experiments with voices synthesized by applying both the source-filter theory and a physical based voice model. The method is also test using human voice signals. The results demonstrate that accurate estimates of the glottal source and the vocal tract filter can be obtained over several scenarios. Moreover, the method is shown to be robust with respect to different phonation types.

A method to estimate glottal source waves and vocal tract shapes for widely pronounced types using ARX-LF model

Accurate estimates of glottal source waves and vocal tract shapes can be used for speech synthesis with high quality. One of source-filter models, ARX-LF model, is widely used to estimate glottal source waves of neutral speech and singing voices. This paper extends the ARX-LF model to accurately estimate glottal source waves and vocal tract shapes with more widely pronounced types. The proposed method mainly contains two steps: the first step is to set initial values of LF parameters and shift glottal closure instant position, and the second step is to save optimum LF parameters and vocal tract coefficients when mean-square equation error is the smallest. The voices are synthesized by different LF parameters and fixed vocal tract filter model, which are used to test whether the proposed method can work well. The results show that the proposed method has ability to estimate LF parameters in more widely pronounced types. The proposed method is further used to estimate glottal source waves and vocal tract sh...

An Investigation on the Accuracy of Truncated DKLT Representation for Speaker Identification With Short Sequences of Speech Frames.

Speaker identification plays a crucial role in biometric person identification as systems based on human speech are increasingly used for the recognition of people. Mel frequency cepstral coefficients (MFCCs) have been widely adopted for decades in speech processing to capture the speech-specific characteristics with a reduced dimensionality. However, although their ability to decorrelate the vocal source and the vocal tract filter make them suitable for speech recognition, they greatly mitigate the speaker variability, a specific characteristic that distinguishes different speakers. This paper presents a theoretical framework and an experimental evaluation showing that reducing the dimension of features by applying the discrete Karhunen-Loève transform (DKLT) to the log-spectrum of the speech signal guarantees better performance compared to conventional MFCC features. In particular with short sequences of speech frames, with typical duration of less than 2 s, the performance of truncated DKLT representation achieved for the identification of five speakers are always better than those achieved with the MFCCs for the experiments we performed. Additionally, the framework was tested on up to 100 TIMIT speakers with sequences of less than 3.5 s showing very good recognition capabilities.

Multi-Frame Amplitude Envelope Estimation for Modification of Singing Voice

Singing voice synthesis benefits from very high quality estimation of the resonances and anti-resonances of the vocal tract filter (VTF), i.e., an amplitude spectral envelope. In the state of the art, a single frame of DFT transform is commonly used as a basis for building spectral envelopes. Even though multiple frame analysis (MFA) has already been suggested for envelope estimation, it is not yet used in concrete applications. Indeed, even though existing attempts have shown very interesting results, we will demonstrate that they are either over complicated or fail to satisfy the high accuracy that is necessary for singing voice. In order to allow future applications of MFA, this article aims to improve the theoretical understanding and advantages of MFA-based methods. The use of singing voice signals is very beneficial for studying MFA methods due to the fact that the VTF configuration can be relatively stable and, at the same time, the vibrato creates a regular variation that is easy to model. By simplifying and extending previous works, we also suggest and describe two MFA-based methods. To better understand the behaviors of the envelope estimates, we designed numerical measurements to assess single frame analysis and MFA methods using synthetic signals. With listening tests, we also designed two proofs of concept using pitch scaling and conversion of timbre. Both evaluations show clear and positive results for MFA-based methods, thus, encouraging this research direction for future applications.

The vocal repertoire of the domesticated zebra finch: a data-driven approach to decipher the information-bearing acoustic features of communication signals.

Although a universal code for the acoustic features of animal vocal communication calls may not exist, the thorough analysis of the distinctive acoustical features of vocalization categories is important not only to decipher the acoustical code for a specific species but also to understand the evolution of communication signals and the mechanisms used to produce and understand them. Here, we recorded more than 8000 examples of almost all the vocalizations of the domesticated zebra finch, Taeniopygia guttata: vocalizations produced to establish contact, to form and maintain pair bonds, to sound an alarm, to communicate distress or to advertise hunger or aggressive intents. We characterized each vocalization type using complete representations that avoided any a priori assumptions on the acoustic code, as well as classical bioacoustics measures that could provide more intuitive interpretations. We then used these acoustical features to rigorously determine the potential information-bearing acoustical features for each vocalization type using both a novel regularized classifier and an unsupervised clustering algorithm. Vocalization categories are discriminated by the shape of their frequency spectrum and by their pitch saliency (noisy to tonal vocalizations) but not particularly by their fundamental frequency. Notably, the spectral shape of zebra finch vocalizations contains peaks or formants that vary systematically across categories and that would be generated by active control of both the vocal organ (source) and the upper vocal tract (filter).

A Time Regularization Technique for Discrete Spectral Envelopes Through Frequency Derivative

In most applications of sinusoidal models for speech signal, an amplitude spectral envelope is necessary. This envelope is not only assumed to fit the vocal tract filter response as accurately as possible, but it should also exhibit slow varying shapes across time. Indeed, time irregularities can generate artifacts in signal manipulations or increase improperly the features variance used in statistical models. In this letter, a simple technique is suggested to improve this time regularity. Considering that time regularity is characterized by slowly varying spectral shapes among successive frames, the basic idea is to smooth the frequency derivative of the envelope instead of its absolute value. Even though, this idea could be applied to different envelope models, the present letter describes its application to the simple linear interpolation envelope. Using real speech signals, the evaluation shows that the time irregularity can be drastically reduced. Additional experiments using synthetic signals also show that the accuracy of the original envelope is not degraded by the process.

Automatic evaluation of hypernasality based on a cleft palate speech database.

The hypernasality is one of the most typical characteristics of cleft palate (CP) speech. The evaluation outcome of hypernasality grading decides the necessity of follow-up surgery. Currently, the evaluation of CP speech is carried out by experienced speech therapists. However, the result strongly depends on their clinical experience and subjective judgment. This work aims to propose an automatic evaluation system for hypernasality grading in CP speech. The database tested in this work is collected by the Hospital of Stomatology, Sichuan University, which has the largest number of CP patients in China. Based on the production process of hypernasality, source sound pulse and vocal tract filter features are presented. These features include pitch, the first and second energy amplified frequency bands, cepstrum based features, MFCC, short-time energy in the sub-bands features. These features combined with KNN classier are applied to automatically classify four grades of hypernasality: normal, mild, moderate and severe. The experiment results show that the proposed system achieves a good performance. The classification rates for four hypernasality grades reach up to 80.4%. The sensitivity of proposed features to the gender is also discussed.

Lingual articulation in songbirds.

Lingual articulation in humans is one of the primary means of vocal tract resonance filtering that produces the characteristic vowel formants of speech. In songbirds, the function of the tongue in song has not been thoroughly examined, although recent research has identified the oropharyngeal-esophageal cavity as a resonance filter that is actively tuned to the frequency of the song. In northern cardinals (Cardinalis cardinalis), the volume of this cavity is inversely proportional to the frequency of the song above 2 kHz. However, cardinal song extends below this range, leaving the question of whether and how the vocal tract is tracking these low frequencies. We investigated the possible role of the tongue in vocal tract filtering using X-ray cineradiography of northern cardinals. Below 2 kHz, there was prominent tongue elevation in which the tip of the tongue was raised until it seemed to touch the palate. These results suggest that tongue elevation lowers the resonance frequency below 2 kHz by reducing the area of the passage from the oral cavity into the beak. This is consistent with a computational model of the songbird vocal tract in which resonance frequencies are actively adjusted by both changing the volume of the oropharyngeal-esophageal cavity and constricting the opening into the beak.

Bandwidth extension of narrowband speech in log spectra domain using neural network

In recent years, there have been significant advances in communication technology, but speech signals still suffer from low perceived quality caused by bandwidth limitations of telephone networks. The bandwidth extension (BWE) approach adds high-frequency components of the speech signal to band-limited telephone speech and increases speech perception significantly. In this work, we develop a new method for representation of vocal tract filter coefficients using log of filter bank energy (LFBE) parameters as an alternative for mel-frequency cepstral coefficients (MFCCs). This approach is based on a strong correlation between the spectral components of low- and high-band spectrums. Furthermore, the performances of Gaussian mixture model and multilayer perceptron neural network methods for estimation of the high-frequency envelope are evaluated. Objective evaluations of the obtained results indicate that the LFBE feature vectors have better performance than the MFCCs. In addition, findings of the objective evaluations showed that using a neural network, which is not common in BWE, achieves a better performance as compared to the Gaussian mixture model.

Nonlinear interactive source-filter models for speech

The linear source-filter model of speech production assumes that the source of the speech sounds is independent of the filter. However, acoustic simulations based on the physical speech production models show that when the fundamental frequency of the source harmonics approaches the first formant of the vocal tract filter, the filter has significant effects on the source due to the nonlinear coupling between them. In this study, two interactive system models are proposed under the quasi steady Bernoulli flow and linear vocal tract assumptions. An algorithm is developed to estimate the model parameters. Glottal flow and the linear vocal tract parameters are found by conventional methods. Rosenberg model is used to synthesize the glottal waveform. A recursive optimization method is proposed to find the parameters of the interactive model. Finally, glottal flow produced by the nonlinear interactive system is computed. The experimental results show that the interactive system model produces fine details of glottal flow source accurately.

Acoustical bases for the perception of simulated laryngeal vocal tremor

Vocal tremor involves atypical modulation of the fundamental frequency (F0) and intensity of the voice. Previous research on vocal tremor has focused on measuring the modulation rate and extent of the F0 and intensity without characterizing other modulations present in the acoustic signal (i.e., modulation of the harmonics). Characteristics of the voice source and vocal tract filter are known to affect the amplitude of the harmonics and could potentially be manipulated to reduce the perception of vocal tremor. The purpose of this study was to determine the adjustments that could be made to the voice source or vocal tract filter to alter the acoustic output and reduce the perception of modulation. This research was carried out using a computational model of speech production that allows for precise control and modulation of the glottal and vocal tract configurations. Results revealed that listeners perceived a higher magnitude of voice modulation when simulated samples had a higher mean F0, greater degree of vocal fold adduction, and vocal tract shape for /i/ vs. /ɑ/. Based on regression analyses, listeners’ judgments were predicted by modulation information present in both low and high frequency bands. [Work supported by NIH F31-DC012697.]

Do subglottal resonances lead to quantal effects resulting in the features [back] and [low]?: A review

A question of general interest is why languages have the sound categories that they do. K. N. Stevens proposed the Quantal Theory of phonological contrasts, suggesting that regions of discontinuity in the articulatory-acoustic mapping serve as category boundaries. H. M. Hanson and K. N. Stevens [Proc. ICPhS, 182–185, 1995] modeled the interaction of subglottal resonances with the vocal-tract filter, showing that when a changing supraglottal formant strays into the territory of a stationary tracheal formant, a discontinuity in supraglottal formant frequency and attenuation of the formant peak occurs. They suggested that vowel space and quality could thus be affected. K. N. Stevens [Acoustic Phonetics, MIT Press, 1998] went further, musing that because the first and second subglottal resonances lead to instabilities in supraglottal formant frequency and amplitude, vowel systems would benefit by avoiding vowels with formants at these frequencies. Avoiding the first subglottal resonance would naturally lead to the division of vowels into those with a low vs. non-low tongue body; avoiding the second would lead to the division of vowels into those having a back vs. front tongue body. We will review subsequent research that offers substantial support for this hypothesis, justifying inclusion of the effects of subglottal resonances in phonological models.

Moving targets and unsteady states: “Shifting” productions of sibilant fricatives by young children

The English voiceless sibilant /s/--/ʃ/ contrast is one that many children do not acquire until their adolescent years. This protracted acquisition may be due to the high level of articulatory control that is necessary to the successful production of an adult-like sibilant, which involves the coordination of lingual, mandibular, and pulmonic gestures. Poor coordination among these gestures can result in the acoustic properties of the noise source or the vocal tract filter changing throughout the timecourse of the frication, to the extent that the phonetic percept of the frication noise changes across its duration. The present study examined such “shifting” productions of sibilant fricatives by native English-acquiring two- through five-year-old children, which were identified from the Paidologos corpus as those productions where the interval of frication was transcribed phonetically as a sequence of fricative sounds. There were two types of shift in frication quality: (1) a gradual change in the resonant frequencies in the spectrogram, suggesting a repositioning of the oral constriction; and (2) an abrupt change in the level of the frication, suggesting a switch in the noise source. Work is underway to develop measures that differentiate these two types of shift, and that suggest their underlying articulatory causes.

Glottal Spectral Separation for Speech Synthesis

This paper proposes an analysis method to separate the glottal source and vocal tract components of speech that is called Glottal Spectral Separation (GSS). This method can produce high-quality synthetic speech using an acoustic glottal source model. In the source-filter models commonly used in speech technology applications it is assumed the source is a spectrally flat excitation signal and the vocal tract filter can be represented by the spectral envelope of speech. Although this model can produce high-quality speech, it has limitations for voice transformation because it does not allow control over glottal parameters which are correlated with voice quality. The main problem with using a speech model that better represents the glottal source and the vocal tract filter is that current analysis methods for separating these components are not robust enough to produce the same speech quality as using a model based on the spectral envelope of speech. The proposed GSS method is an attempt to overcome this problem, and consists of the following three steps. Initially, the glottal source signal is estimated from the speech signal. Then, the speech spectrum is divided by the spectral envelope of the glottal source signal in order to remove the glottal source effects from the speech signal. Finally, the vocal tract transfer function is obtained by computing the spectral envelope of the resulting signal. In this work, the glottal source signal is represented using the Liljencrants-Fant model (LF-model). The experiments we present here show that the analysis-synthesis technique based on GSS can produce speech comparable to that of a high-quality vocoder that is based on the spectral envelope representation. However, it also permit control over voice qualities, namely to transform a modal voice into breathy and tense, by modifying the glottal parameters.

Analysis and Recognition of Vowels in SHAI`YÂNG MIRI Language using Formants

Speech is generated by the co-ordination of various anatomical articulations known collectively as the vocal organs. The waveforms thus produced are either voiced or unvoiced speech. All the vowel sounds are voiced and have the sound source, the glottis. An amplification caused by the vocal tract filter is called resonance, and in speech these resonance are known as formants. Since each vowel has a different vocal tract shape, it will have a different format pattern. This paper is an attempt to analyze the vowels in SHAI`YÂNG MIRI language using the formant analysis. Analysis of the eight cardinal vowel speech samples in SHAI`YÂNG MIRI language is performed using formants, spectrogram, 19-channel auditory filter bank and through normalization of vowels. General Terms Speech Processing and Analysis, Speech Synthesis

Automatic glottal inverse filtering with the Markov chain Monte Carlo method

This paper presents a new glottal inverse filtering (GIF) method that utilizes a Markov chain Monte Carlo (MCMC) algorithm. First, initial estimates of the vocal tract and glottal flow are evaluated by an existing GIF method, iterative adaptive inverse filtering (IAIF). Simultaneously, the initially estimated glottal flow is synthesized using the Rosenberg–Klatt (RK) model and filtered with the estimated vocal tract filter to create a synthetic speech frame. In the MCMC estimation process, the first few poles of the initial vocal tract model and the RK excitation parameter are refined in order to minimize the error between the synthetic and original speech signals in the time and frequency domain. MCMC approximates the posterior distribution of the parameters, and the final estimate of the vocal tract is found by averaging the parameter values of the Markov chain. Experiments with synthetic vowels produced by a physical modeling approach show that the MCMC-based GIF method gives more accurate results compared to two known reference methods.

Spectrographic analysis points to source–filter coupling in rutting roars of Iberian red deer

Source–filter coupling is the rarest acoustic phenomenon not only in Iberian red deer, but in any mammal. In most mammals, sound production can be well described in the framework of source–filter theory. The vocal output is the result of combined work of the larynx (the source) and of the supralaryngeal vocal tract (the filter). The source–filter theory suggests the independence of source and filter. Thus, vocal tract filtering should not affect the fundamental frequency (f0) of the sound created in the larynx. Spectrographically, the source is mostly characterized by the f0 and its harmonics, while the filter by the vocal tract resonances, i.e., formant frequencies. Nevertheless, a non-independent (coupled) source and filter can be proposed when the vocal folds start oscillating at one of the formant frequencies. Coupling between source and filter has been found in human singers and predicted for red deer Cervus elaphus by a computer modeling approach. This study describes different modes of phonation in a natural bout of rutting calls of Iberian red deer Cervus elaphus hispanicus and the transition from a chaotic mode to a probable source–filter coupling mode. This phenomenon might be involved in the production of extremely high-frequency bugles of North American and Asian subspecies of C. elaphus.

Mixed source model and its adapted vocal tract filter estimate for voice transformation and synthesis

In current methods for voice transformation and speech synthesis, the vocal tract filter is usually assumed to be excited by a flat amplitude spectrum. In this article, we present a method using a mixed source model defined as a mixture of the Liljencrants–Fant (LF) model and Gaussian noise. Using the LF model, the base approach used in this presented work is therefore close to a vocoder using exogenous input like ARX-based methods or the Glottal Spectral Separation (GSS) method. Such approaches are therefore dedicated to voice processing promising an improved naturalness compared to generic signal models. To estimate the Vocal Tract Filter (VTF), using spectral division like in GSS, we show that a glottal source model can be used with any envelope estimation method conversely to ARX approach where a least square AR solution is used. We therefore derive a VTF estimate which takes into account the amplitude spectra of both deterministic and random components of the glottal source. The proposed mixed source model is controlled by a small set of intuitive and independent parameters. The relevance of this voice production model is evaluated, through listening tests, in the context of resynthesis, HMM-based speech synthesis, breathiness modification and pitch transposition.

감정 인식을 위한 음성 특징 도출

Emotion recognition is an important technology in the filed of human-machine interface. To apply speech technology to emotion recognition, this study aims to establish a relationship between emotional groups and their corresponding voice characteristics by investigating various speech features. The speech features related to speech source and vocal tract filter are included. Experimental results show that statistically significant speech parameters for classifying the emotional groups are mainly related to speech sources such as jitter, shimmer, F0 (F0_min, F0_max, F0_mean, F0_std), harmonic parameters (H1, H2, HNR05, HNR15, HNR25, HNR35), and SPI.