LPC Analysis Research Articles

Mechanical models as ground truth for vowel resonance analysis

It has long been established that LPC analysis results in formant estimates that are not accurate representations of the resonances; they are biased towards the nearest harmonic, and this bias worsens as F0 rises to 200 Hz or more. Manual measurement of formants with the reassigned spectrogram (RS) has been shown to be more accurate, but the ground truth is needed in order to test automatic measurement methods. Here, “vocal tract” models were 3D-printed to allow resonances to be excited replicably. Using the principle of acoustic reciprocity, source signals were played over an external loudspeaker, and the output filtered by the model was recorded via a microphone located in the model’s “glottis.” A white noise source signal was used to determine the resonances. Sawtooth and impulse train source signals (F0s 83–400 Hz) were filtered by the physical model and then analyzed with both LPC and RS. LPC formants were biased towards the nearest harmonic, resulting in errors from 10 Hz at low F0 to 250 at 300 Hz. RS errors ranged from 0 to 20 Hz and were not correlated with F0. These results indicate physical models are useful for creating known resonances for validating accuracy in our measurements.

A NOVEL APPROACH TO STUTTERED SPEECH CORRECTION

Stuttered speech is a dysfluency rich speech, more prevalent in males than females. It has been associated with insufficient air pressure or poor articulation, even though the root causes are more complex. The primary features include prolonged speech and repetitive speech, while some of its secondary features include, anxiety, fear, and shame. This study used LPC analysis and synthesis algorithms to reconstruct the stuttered speech. The results were evaluated using cepstral distance, Itakura-Saito distance, mean square error, and likelihood ratio. These measures implied perfect speech reconstruction quality. ASR was used for further testing, and the results showed that all the reconstructed speech samples were perfectly recognized while only three samples of the original speech were perfectly recognized.

The recorder's paradox: Balancing high-quality recordings with spontaneous speech in noisy recording environments

While sociophonetic analysis requires high-quality sound recordings, sociolinguistic interviews (Labov 1984) are often conducted in uncontrolled, natural environments to elicit casual speech (Tagliamonte 2006). The effects of room acoustics and background noise on formant measurements, however, have never been systematically examined. To empirically investigate how ambient noise affects recording quality and measurements, a male speaker of English was simultaneously recorded to multiple devices reading 260 carrier phrases. Three naturally occurring background noise conditions ( + 20dB, + 10dB, and + 0dB SNR) were created in Praat 5.4 (Boersma and Weenink 2014) and mixed with the original audio recordings. 10,028 measurements of F1 and F2 were taken at the temporal midpoint of each vowel in each of the four recordings using LPC analysis in Praat. Pearson's r tests in R (R Core Team, 2014) assessed the correlation between measurements from each recording. Main results reveal positive correlations between “noiseless” and + 20dB, and + 10dB SNR conditions for each device, all vowels and both formants. When the signal was not appreciably louder than the background noise (i.e. + 0 dB SNR) correlations significantly weakened. These findings are discussed as they relate to sociolinguistic investigations that need to balance high-fidelity recordings with noisier speaking environments.

Validation of Optimum Algorithm Parameters Required to Estimate Vocal Tract Shape for Children Using LPC Analysis

Severe or profound deafness in hearing impaired children, can curb their ability to speak due to the lack of auditory feedback. There has been a considerable attempt in developing commercial speech training aids for such children which give feedback of acoustic and articulatory parameters. Speech training aids based on visual feedback of vocal tract shape (VTS) are reported to be useful for the improvement in speech production. Since realistic VTS estimation for adult speakers and their validation has already been done successfully, VTS estimation is now necessarily required in case of children too, so that they get trained in speech at an early age. The investigation on vocal tract shape estimation based on LPC analysis of speech by appropriately selecting some of the algorithm parameters such as vocal tract length, LPC order, and speech sampling rate has been done in our previous work. This paper attempts to validate the obtained results for vocal tract shapes corresponding to certain recorded vowels from children belonging to specific age groups. Since MRI images of VTS are unavailable for articulating children, validation of our results is based on the results from researchers who have used other indirect techniques to obtain VTS.

Functional Properties of Novel Metallic Hard Masks

Hard masks play an important role in pattern transfer to the desired substrate in the semiconductor lithography processes. Organic and inorganic type hard masks are used. While most organic hard masks such as carbon and siloxane type are solution spin coated, inorganic type hard masks such as SiON and SiN are either chemical vapor deposited (CVD) or atomic layer deposited (ALD). Future generation of lithography processes require hard masks with higher resistance to fluorinated plasma and materials that can be easily wet stripped after pattern transfer process to prevent dry etch damage to the substrate underneath.The present paper describes formation and functional properties of novel metal oxide hard masks by simple solution spin coating process. These novel metal oxide hard masks offer good etch selectivity and can easily be partially or fully wet strippable using commonly used chemicals in the FABs. The spin coatable composition has good long-term shelf life and pot life stability based on solution LPC analysis and wafer defect studies. The hard mask material absorbs DUV wavelengths and can be used as a spin-on inorganic or hybrid antireflective coating to control substrate reflectivity under DUV exposure of photoresist. At the same time they are transparent at 500-700nm for alignment mark identification and can be spin coatable up to 450nm thickness with good film quality. Some of these metal-containing materials can be used as an underlayer in EUV lithography to significantly enhance sensitivity of the photoresist. Specific metal hard masks are also developed for via or trench filling applications in IRT processes. The materials have shown good coating and lithography performance with a film thicknesses as low as 10 nm under ArF dry or immersion conditions.

ISOLATED WORD RECOGNITION USING LPC & VECTOR QUANTIZATION

Speech recognition is always looked upon as a fascinating field in human computer interaction. It is one of the fundamental steps towards understanding human recognition and their behavior. This paper explicates the theory and implementation of Speech recognition. This is a speaker-dependent real time isolated word recognizer. The major logic used was to first obtain the feature vectors using LPC which was followed by vector quantization. The quantized vectors were then recognized by measuring the Minimum average distortion. All Speech Recognition systems contain Two Main Phases, namely Training Phase and Testing Phase. In the Training Phase, the Features of the words are extracted and during the recognition phase feature matching Takes place. The feature or the template thus extracted is stored in the data base, during the recognition phase the extracted features are compared with the template in the database. The features of the words are extracted by using LPC analysis. Vector Quantization is used for generating the code books. Finally the recognition decision is made based on the matching score. MATLAB will be used to implement this concept to achieve further understanding.

Developmental changes of nasal and oral calls in the goitred gazelle Gazella subgutturosa, a nonhuman mammal with a sexually dimorphic and descended larynx

In goitred gazelles (Gazella subgutturosa), sexual dimorphism of larynx size and position is reminiscent of the case in humans, suggesting shared features of vocal ontogenesis in both species. This study investigates the ontogeny of nasal and oral calls in 23 (10 male and 13 female) individually identified goitred gazelles from shortly after birth up to adolescence. The fundamental frequency (f0) and formants were measured as the acoustic correlates of the developing sexual dimorphism. Settings for LPC analysis of formants were based on anatomical dissections of 5 specimens. Along ontogenesis, compared to females, male f0 was consistently lower both in oral and nasal calls and male formants were lower in oral calls, whereas the first two formants of nasal calls did not differ between sexes. In goitred gazelles, significant sex differences in f0 and formants appeared as early as the second week of life, while in humans they emerge only before puberty. This result suggests different pathways of vocal ontogenesis in the goitred gazelles and in humans.

Modulation-domain Kalman filtering for single-channel speech enhancement

In this paper, we investigate the modulation-domain Kalman filter (MDKF) and compare its performance with other time-domain and acoustic-domain speech enhancement methods. In contrast to previously reported modulation domain-enhancement methods based on fixed bandpass filtering, the MDKF is an adaptive and linear MMSE estimator that uses models of the temporal changes of the magnitude spectrum for both speech and noise. Also, because the Kalman filter is a joint magnitude and phase spectrum estimator, under non-stationarity assumptions, it is highly suited for modulation-domain processing, as phase information has been shown to play an important role in the modulation domain. We have found that the Kalman filter is better suited for processing in the modulation-domain, rather than in the time-domain, since the low order linear predictor is sufficient at modelling the dynamics of slow changes in the modulation domain, while being insufficient at modelling the long-term correlation speech information in the time domain. As a result, the MDKF method produces enhanced speech that has very minimal distortion and residual noise, in the ideal case. The results from objective experiments and blind subjective listening tests using the NOIZEUS corpus show that the MDKF (with clean speech parameters) outperforms all the acoustic and time-domain enhancement methods that were evaluated, including the time-domain Kalman filter with clean speech parameters. A practical MDKF that uses the MMSE-STSA method to enhance noisy speech in the acoustic domain prior to LPC analysis was also evaluated and showed promising results.

Method and Apparatus for Pitch Modulation and Gender Identification of a Voice Signal

A method and apparatus for modulating pitch and identifying gender of a voice signal is provided. The voice modulation apparatus includes an LPC analyzer for obtaining vocal track filter coefficients reflecting characteristics of an input voice signal, a pitch detector for detecting the pitch and the gender of the voice signal, a pitch modulator for modulating the voice signal by applying a predetermined value to a detected value from the pitch detector, and a coder for coding the input signal from to LPC analyzer and to pitch modulator, and for outputting a coded signal.

Empirical articulatory‐acoustic relations for vowels

Vowels tokens were extracted from four talkers in the Wisconsin X‐ray Microbeam Speech Production Database. The neighboring phonemes of these vowels were restricted to be non‐nasal and non‐liquid. The first three formant frequencies were measured using LPC analysis with manual corrections at a rate corresponding to the pellet trajectory sampling rate, thus yielding large amounts of simultaneous formant frequency and pellet position data points (between 11,000 and 20,000 for each talker.) Principal components analysis was performed for both the formant frequencies and the pellet position data, to produce three orthogonal acoustic components and four orthogonal articulatory components. A local linear regression technique, known as loess [Cleveland, W. S. and Devlin, S. J. (1988), J. Amer. Stat. Assoc., 83, 596 ‐ 610], was applied to orthogonal components to map between the acoustic and articulatory domains. This technique permits regression slopes to vary within the domain of the independent variables. The results will be discussed in terms of optimization of loess parameters (e.g. size of local neighborhoods), goodness of fit of the mappings, and the degree to which slopes in the mappings vary. Manually corrected formant frequencies will be compared with fully automatic Line Spectral Frequencies. [Supported by NIDCD‐001247 to CReSS LLC]

Robust F0 Estimation Using ELS-Based Robust Complex Speech Analysis

Complex speech analysis for an analytic speech signal can accurately estimate the spectrum in low frequencies since the analytic signal provides spectrum only over positive frequencies. The remarkable feature makes it possible to realize more accurate F0 estimation using complex residual signal extracted by complex-valued speech analysis. We have already proposed F0 estimation using complex LPC residual, in which the autocorrelation function weighted by AMDF was adopted as the criterion. The method adopted MMSE-based complex LPC analysis and it has been reported that it can estimate more accurate F0 for IRS filtered speech corrupted by white Gauss noise although it can not work better for the IRS filtered speech corrupted by pink noise. In this paper, robust complex speech analysis based on ELS (Extended Least Square) method is introduced in order to overcome the drawback. The experimental results for additive white Gauss or pink noise demonstrate that the proposed algorithm based on robust ELS-based complex AR analysis can perform better than other methods.

Robust F0 Estimation Based on Complex LPC Analysis for IRS Filtered Noisy Speech

This paper proposes a novel robust fundamental frequency (F0) estimation algorithm based on complex-valued speech analysis for an analytic speech signal. Since analytic signal provides spectra only over positive frequencies, spectra can be accurately estimated in low frequencies. Consequently, it is considered that F0 estimation using the residual signal extracted by complex-valued speech analysis can perform better for F0 estimation than that for the residual signal extracted by conventional real-valued LPC analysis. In this paper, the autocorrelation function weighted by AMDF is adopted for the F0 estimation criterion and four signals; speech signal, analytic speech signal, LPC residual and complex LPC residual, are evaluated for the F0 estimation. Speech signals used in the experiments were an IRS filtered speech corrupted by adding white Gaussian noise or Pink noise whose noise levels are 10, 5, 0, -5[dB]. The experimental results demonstrate that the proposed algorithm based on complex LPC residual can perform better than other methods in noisy environment.

A low-bit-rate audio codec using mel-scaled linear predictive analysis

In this paper, we propose a low-bit-rate audio codec using a new analysis method named mel-scaled linear predictive analysis (mel-LP analysis). In mel-LP analysis, a spectral envelope is estimated on a mel- or bark-frequency scale, so as to improve the spectral resolution in the low-frequency band. This analysis is accomplished with about a twofold increase in computation over standard LPC analysis. Our codec using mel-LP analysis consists of five key parts: time frequency transformation, flattening of MDCT coefficients using the mel-LP spectral envelope, power normalization, perceptual weighting estimation, and multistage VQ. In subjective experiments, we investigated the performance of our codec using the mel-LP analysis method, through 7-level paired comparison tests. The result shows that the codec using the mel-LP analysis method results in a good performance at a low bit rate, particularly at 16kbps. In the cases of pop songs, piano music and male speech, sound quality was improved.

A noise robust feature extraction algorithm using joint wavelet packet subband decomposition and AR modeling of speech signals

This paper presents a noise robust feature extraction algorithm NRFE using joint wavelet packet decomposition (WPD) and autoregressive (AR) modeling of a speech signal. In opposition to the short time Fourier transform (STFT)-based time–frequency signal representation, wavelet packet decomposition can lead to better representation of non-stationary parts of the speech signal (e.g. consonants). The vowels are well described with an AR model as in LPC analysis. The proposed Root-Log compression scheme is used to perform the computation of the wavelet packet parameters. The separately extracted WPD and AR-based parameters are combined together and then transformed with the usage of linear discriminant analysis (LDA) to finally produce a lower dimensional output feature vector. The noise robustness is improved with the application of proposed wavelet-based denoising algorithm with a modified soft thresholding procedure and time–frequency adaptive threshold. The proposed voice activity detector based on a skewness-to-kurtosis ratio of the LPC residual signal is used to effectively perform a frame-dropping principle. The speech recognition results achieved on Aurora 2 and Aurora 3 databases show overall performance improvement of 44.7% and 48.2% relative to the baseline MFCC front-end, respectively.

Interpolation of vocal-tract shape during stop closures from transition segments in vowel-consonant-vowel syllables

Children with prelingual profound hearing impairment have great difficulty in acquiring speech. Speech-training systems providing visual feedback of vocal-tract shape are found to be useful for improving vowel articulation. Vocal-tract shape estimation, based on LPC and other analysis techniques, generally fails during stop closures, and this restricts its effectiveness in speech training for production of consonants not having visible articulatory efforts. A technique based on two-dimensional surface modeling of the area values, estimated by LPC analysis, during the vowel-consonant and consonant-vowel transitions preceding and following the stop closure, has been investigated for interpolating the area values during the stop closures. Surface modeling was based on least-squares bivariate polynomials and Delaunay triangulation methods. Syllables of the type /aCa/, /aCi/, /iCa/, /iCi/, and /uCu/ with stop consonants /p/, /b/, /t/, /d/, /k/, and /g/ were analyzed for the estimation of place of stop closure. For bilabial, alveolar, and velar stops, the place could be estimated consistently with conic polynomial surface interpolation. Estimation of place, based on Delaunay surfaces, was consistent for bilabial and velar stops. However, cubic polynomial surface interpolation results were less consistent in estimating the place of constriction.

Robust F0 estimation based on complex speech analysis

This paper proposes a novel robust fundamental frequency (F0) estimation algorithm based on complex‐valued speech analysis for an analytic speech signal. Since the analytic signal provides spectra only over positive frequencies, spectra can be estimated accurately at low frequencies. Consequently, it is considered that F0 estimation using the residual signal extracted by complex‐valued speech analysis can perform better for F0 estimation than that for the residual signal extracted by conventional real‐valued LPC analysis. In this paper, the autocorrelation function (AUTOC) weighted by a reciprocal of the AMDF is also adopted for the F0 estimation criterion. The proposed F0 estimation algorithm is evaluated using three criteria, AUTOC, AMDF, and the weighted AUTOC, with complex‐valued residual extracted by complex‐valued speech analysis. We also compared the proposed method with those for three signals: the speech signal, analytic speech signal, and LPC residual. Speech signals used in the experiments were corrupted by addition of white Gaussian noise whose noise levels are 10, 5, 0, and −5 [dB]. The performance was evaluated using the gross pitch error (GPE) and fine pitch error (FPE). The results demonstrate that this algorithm based on complex speech analysis can perform better than conventional methods in an extremely noisy environment.

Lip protrusion and tongue position in French vowels produced by blind speakers and sighted speakers

The influence of visual experience in speech production was investigated through a study of the French vowels /i y u a/. Articulatory and acoustic recordings of ten repetitions of the vowels in CVC syllables embedded in carrier sentences were conducted in three consonant environments (/b d g/) and two prosodic conditions (neutral and under contrastive emphasis). Six congenitally blind adults and six sighted adults were recorded. All participants were native speakers of Canadian French and had no history of speech disorder. The audio signal, lip movement, and tongue shapes were recorded using a digital camera and an ultrasound system. The minimal (for /i a/) or maximal (for /y u/) horizontal position of the upper lip was tracked and the front-back position of the tongue was measured for each vowel. Formant frequencies were extracted using LPC analyses. Results show that, despite similar acoustic differences, the rounded and unrounded vowels are less differentiated along the protrusion dimension for blind speakers compared to sighted speakers. It is suggested that this variability is related to a trade-off between lip protrusion and tongue position. Significant interaction effects between consonant environment and vowel category are discussed for each speaker group. [Work supported by SSHRC, NSERC.]

Knowledge-based formant tracking with confidence measure using dynamic programming

In this study, we report on refinements to a formant tracking technique originally reported in [Xia et al., ICSLP (2000)]. The formant tracker operates in two phases. First, it finds optimal formant track estimates in oral sonorant regions by imposing frequency continuity constraints using dynamic programming. Second, post-processing is performed to make the estimates more robust and accurate, and to extend formant tracks in nasal and obstruent regions. In recent work, we have improved on our initial estimates of the formants by combining the outputs from a 12th-order LPC analysis and a 16th-order LPC analysis. Additionally, we have added a confidence measure for each formant track in each frame. The confidence measure is based on formant continuity, competing formants, short-time energy, context information, and formant information over the entire utterance. The experiments show that most of the tracking errors are associated with a low confidence value, while the correct formants have high confidence values. The performance of the algorithm in the sonorant regions was tested using randomly selected male and female speech from the TIMIT database. [Work supported by NSF Grant No. BCS0236707.]

Effect of multiple compression channels on vowel spectra

Multichannel wide-dynamic range compression (MCC) is a common amplification strategy, but whether using more channels provides an advantage for speech recognition is disputed. Some researchers have hypothesized that MCC with a large number of compression channels can reduce spectral contrast and degrade recognition, but this has not been empirically tested. A detailed acoustic analysis was undertaken to determine (1) whether vowel spectra are altered by multichannel compression relative to single-channel compression, and (2) to quantify changes to vowel spectra as a function of the number of compression channels. Stimuli were eight naturally spoken vowels produced by five males and five females in an /hVd/ context. A series of compression conditions (1 through 16 channels) was created using clinically representative parameters. After compression processing, F1, F2, and F3 were measured using spectrograms, LPC, and FFT analysis. A quantitative measure was developed to represent spectral changes after compression. Preliminary data show spectral smoothing with more than four channels. Smaller numbers of compression channels have minimal effects on vowel spectra. These data will be related to vowel recognition scores.

스펙트럼 형태 불변 실시간 음성 변환 시스템

본 논문에서는 음성의 스펙트럼 형태는 유지하면서 음성을 기계적인 음성으로 변환시키기는 실시간 음성 변환 방법을 제안하였다. 이러한 목적을 위하여 LPC 분석 및 합성 방법을 사용하여 변환된 음성의 스펙트럼은 유지하였고 합성된 음성의 피치는 자유롭게 변경되도록 하였다. 제안된 방법에서는 변환된 음성이 보다 자연스럽게 들리게 하기 위하여 여기 신호 발생기에 이득 정합 방법을 적용하였다. 제안된 방법의 성능을 평가하기 위하여 음성 변환 실험을 수행하였다. 실험 결과에서 원 음성 신호는 원 화자의 신원을 알기가 어려운 기계적인 음성 신호로 바뀌는 것을 알 수 있었고 피치의 심한 변화에도 변환된 음성의 의미는 정확히 전달될 수 있었다. 제안된 시스템은 시스템의 실시간으로 구현될 수 있는지 확인하기 위하여 TI TMS320C6711DSK 보드를 사용하여 구현되었다. In this paper, the spectral shape invariant real-time voice change method is proposed to change one's voice to mechanical voice. For this purpose, LPC analysis and synthesis is used to maintain the spectraum of voice and the pitch of synthesis speech can be changed freely. In the proposed method, gain matching method is applied to excitation signal generator to make the changed voice natural to hear. In order to evaluate the performance of the proposed method, voice change experiments were conducted. Experimental results showed that original speech signal is changed to the mechanical voice signal in which context of the speaker's voice is conveyed correctly in spite of drastic change of pitch. The system is implemented using TI TMS320C6711DSK board to verify the system runs in real time.