Linear Prediction Spectrum Research Articles

Linear prediction on Cent scale for fundamental frequency analysis

Understanding the fundamental frequency and harmonic content of audio signals is crucial for many applications in music analysis, including music transcription, audio synthesis, and genre identification. This study formulates a signal processing approach combining Linear Prediction (LP) analysis and the Cent scale to accurately characterize the pitch and harmonic structure of the audio signals. Pitch tracking on the LP spectrum in the Cent scale provides more accurate and reliable pitch estimation, especially in the presence of noise or overlapping harmonics. The Cent scale aligns the harmonics of different notes more closely, making it easier to discern the correct pitch.

Power spectrum analysis method of coherent Doppler lidar based on linear prediction spectrum estimation

Power spectrum analysis method of coherent Doppler lidar based on linear prediction spectrum estimation

Data Augmentation Using Spectral Warping for Low Resource Children ASR

In low resource children automatic speech recognition (ASR) the performance is degraded due to limited acoustic and speaker variability available in small datasets. In this paper, we propose a spectral warping based data augmentation method to capture more acoustic and speaker variability. This is carried out by warping the linear prediction (LP) spectra computed from speech data. The warped LP spectra computed in a frame-based manner are used with the corresponding LP residuals to synthesize speech to capture more variability. The proposed augmentation method is shown to improve the ASR system performance over the baseline system. We have compared the proposed method with four well-known data augmentation methods: pitch scaling, speaking rate, SpecAug and vocal tract length perturbation (VTLP), and found that the proposed method performs the best. Further, we have combined the proposed method with these existing data augmentation methods to improve the ASR system performance even more. The combined system consisting of the original data, VTLP, SpecAug and the proposed spectral warping method gave the best performance by a relative word error rate reduction of 32.13% and 10.51% over the baseline system for Punjabi children and TLT-school corpus, respectively. The proposed spectral warping method is publicly available at https://github.com/kathania/Spectral-Warping.

A formant modification method for improved ASR of children’s speech

Differences in acoustic characteristics between children’s and adults’ speech degrade performance of automatic speech recognition systems when systems trained using adults’ speech are used to recognize children’s speech. This performance degradation is due to the acoustic mismatch between training and testing. One of the main sources of the acoustic mismatch is the difference in vocal tract resonances (formant frequencies) between adult and child speakers. The present study aims to reduce the mismatch in formant frequencies by modifying formants of children’s speech to better correspond to formants of adults’ speech. This is carried out by warping the linear prediction (LP) spectrum computed from children’s speech. The warped LP spectra computed in a frame-based manner from children’s speech are used with the corresponding LP residuals to synthesize speech whose formant structure is closer to that of adults’ speech. When used in testing of an ASR system trained using adults’ speech, the warping reduces the spectral mismatch in speech between training and testing and improves the system performance in recognition of children’s speech. Experiments were conducted using narrowband (8 kHz) and wideband (16 kHz) speech of adult and child speakers from the WSJCAM0 and PF_STAR databases, respectively, and by recognizing children’s speech using acoustic models trained with adults’ speech. The proposed method gave relative improvements of 24% and 11% for the DNN and TDNN acoustic models, respectively, for narrowband speech. For wideband speech, the technique gave relative improvements of 27% and 13% for the DNN and TDNN acoustic models, respectively. The performance of the proposed method was also compared to two speaker adaptation methods: vocal tract length normalization (VTLN) and speaking rate adaptation (SRA). This comparison showed the best recognition performance for the proposed method. We also combined the proposed method with VTLN and SRA, and found that the combined method gave a further reduction in WER. Moreover, our experiments carried out for noisy speech using various types of additive noise and signal-to-noise ratios showed that the proposed method performs well also for degraded speech.

Analysis of Articulation Errors in Dysarthric Speech.

Imprecise articulation is the major issue reported in various types of dysarthria. Detection of articulation errors can help in diagnosis. The cues derived from both the burst and the formant transitions contribute to the discrimination of place of articulation of stops. It is believed that any acoustic deviations in stops due to articulation error can be analyzed by deriving features around the burst and the voicing onsets. The derived features can be used to discriminate the normal and dysarthric speech. In this work, a method is proposed to differentiate the voiceless stops produced by the normal speakers from the dysarthric by deriving the spectral moments, two-dimensional discrete cosine transform of linear prediction spectrum and Mel frequency cepstral coefficients features. These features and cosine distance based classifier is used for the classification of normal and dysarthic speech.

Word recognition from speech signal using linear predictive coding and spectrum analysis

This paper presents a technique for isolated word recognition from speech signal using Spectrum Analysis and Linear Predictive Coding (LPC). In the present study, only those words have been analyzed which are commonly used during a telephonic conversations by criminals. Since each word is characterized by unique frequency spectrum signature, thus, spectrum analysis of a speech signal has been done using certain statistical parameters. These parameters help in recognizing a particular word from a speech signal, as there is a unique value of a feature for each word, which helps in distinguishing one word from the other. Second method used is based on LPC coefficients. Analysis of features extracted using LPC coefficients help in identification of a specific word from the input speech signal. Finally, a combination of best features from these two methods has been used and a hybrid technique is proposed. An accuracy of 94% has been achieved for sample size of 400 speech words.

Automated formant tracking using reassigned spectrograms

The measurement and tracking of formant frequencies is usually accomplished with a linear prediction spectrum estimate, combined with a heuristic-based tracking algorithm. Despite the broad acceptance of this procedure over decades, there is a known bias towards the nearest harmonic, and formant estimates are less accurate for speech with higher fundamental frequency. Previous research has established the superior accuracy of formant measurement using the reassigned spectrogram; however, the measurements must be done by hand, severely limiting its usefulness. Here we present a technique that can locate formants automatically in a reassigned spectrogram. First, the reassigned spectrogram of the signal must be heavily “pruned” so that only a few points remain which highlight the detected frequency components, which are the presumed formants. Second, we apply a simple ridge-finding routine to the pruned spectrogram, to determine the formant tracks automatically. This process was tested on speech samples—/hVd/-words—by four speakers previously used in a comparison of 5 automatic algorithms and manually-measured reassigned spectrograms, and on breathy and falling tones in a speaker of Hmong. The results of this process are generally as good or better than the manual measurement when the optimal parameters for ridge-finding are used.

A Smoothed and Thresholded Linear Prediction Analysis for Efficient Speech Coding

In this paper, we propose a new method of linear prediction (LP) analysis for estimating LP coefficients that are used in current speech coders. This method improves the robustness of LP coefficients computed from speech signals corrupted by noise as well as offering better quantisation efficiency. The quantisation performance and the noise-robustness obtained by the proposed LP coefficients were compared to that obtained by the LP coefficients computed using the autocorrelation and LP spectrum modification methods of LP Analysis, in terms of spectral distortion (SD). The results indicate that the proposed LP coefficients were more robust to noise and also offered transparent quantisation at lower bit-rates (savings of up to 2 bits/frame) than other LP coefficients.

정상 및 심질환 소아의 청진음 분석에 관한 연구

This study tries to analyze morphology and formant frequencies of linear prediction spectra of stethoscope sounds for heart diseased children. For this object, heart diseased stethoscope sounds were collected in the pediatrics of an university hospital. The collected signals were preprocessed and analyzed by the Burg algorithm, a kind of linear prediction analysis. The linear prediction spectra and the formant frequencies of the spectra for the stethoscope sounds for the normal and the diseased children are estimated and compared. The spectra showed outstanding differences in morphology and formant frequencies between the normal and the diseased children. Normal children showed relatively low frequency of F1(the first formant) and small negative slope from F1. VSD children revealed stiff slope change around F1 to F3. Spectra of ASD children is similar with the normal case, but have negative values of F3. F1-F2 difference of the functional murmur children were relatively large.

심음도 스펙트럼의 1, 2차 도함수를 이용한 형성음 주파수 추출 기술

This study describes a new method to analyze phonocardiogram acquired from electronic stethoscope. The method uses the formant frequencies of linear prediction spectrum of the phonocardiogram and proposes a novel method for formant detection using the smoothing and the first and second derivatives. For this, stethoscope sounds are acquired in university hospital. The stethoscope signals are preprocessed and analyzed by the Burg algorithm, a kind of linear prediction analysis. Based on the linear prediction spectra, the formant frequencies are estimated. The proposed method has shown better performance in formant frequency detection than the conventional peak picking method.

Pitch-Scaled Spectrum Based Excitation Model for HMM-based Speech Synthesis

The speech generated by hidden Markov model (HMM)-based speech synthesis systems (HTS) suffers from a ‘buzzing’ sound, which is due to an over-simplified vocoding technique. This paper proposes a new excitation model that uses a pitch-scaled spectrum for the parametric representation of speech in HTS. A residual signal produced using inverse filtering retains the detailed harmonic structure of speech that is not part of the linear prediction (LP) spectrum. By using pitch-scaled spectrums, we can compensate the LP spectrum using the detailed harmonic structure of the residual signal. This spectrum can be compressed using a periodic excitation parameter so that it can used to train HTS. We define an aperiodic measure as the harmonics-to-noise ratio, and calculate a voicing-cut off frequency to fit the aperiodic measure to a sigmoid function. We combine the LP coefficient, pitch-scaled spectrum, and sigmoid function to create a new parametric representation of speech. Listening tests were carried out to evaluate the effectiveness of the proposed technique. This vocoder received a mean opinion score of 4.0 in analysis-synthesis experiments, before dimensionality reduction. By integrating this vocoder into HTS, we improved the sound of the synthesized speech compared with the pulse train excitation model, and demonstrated an even better result than STRAIGHT-HTS.

LP spectra vs. Mel spectra for identification of professional mimics in Indian languages

Automatic Speaker Recognition (ASR) is an economic tool for voice biometrics because of availability of low cost and powerful processors. For an ASR system to be successful in practical environments, it must have high mimic resistance, i.e., the system should not be defeated by determined mimics which may be either identical twins or professional mimics. In this paper, we demonstrate the effectiveness of Linear Prediction (LP)-based features, viz., Linear Prediction Coefficients (LPC) and Linear Prediction Cepstral Coefficients (LPCC) over filterbank-based features such as Mel-Frequency Cepstral Coefficients (MFCC) and newly proposed Teager energy-based MFCC (T-MFCC) for the identification of professional mimics in Indian languages. Results are reported for real and fictitious experiments. On the whole, it is observed that LP-based features perform better than filterbank-based features (an average jump of 23.21% and 31.43% for fictitious experiments with professional mimic in Marathi and Hindi, respectively, whereas there is an average jump of 1.64% for real experiments with professional mimic in Hindi) and we believe that this is the first time such results on identification of professional mimics in ASR are obtained. Analysis of the results is given with the help of Mean Square Error (MSE) between training and testing utterances for mimic’s imitations for target speakers and target speakers’ normal voice. Fourier spectra and corresponding LP spectra for target speaker and its impersonations provided by professional mimic are shown to justify the results. Finally, dependence of LPC on physiological characteristics of vocal tract and its relation with respect to the problem addressed in this paper is studied.

Modification of pitch using DCT in the source domain

In this paper, we propose a novel algorithm for pitch modification. The linear prediction residual is obtained from pitch synchronous frames by inverse filtering the speech signal. Then the discrete cosine transform (DCT) of these residual frames is taken. Based on the desired factor of pitch modification, the dimension of the DCT coefficients of the residual is modified by truncating or zero padding, and then the inverse discrete cosine transform is obtained. This period modified residual signal is then forward filtered to obtain the pitch modified speech. The mismatch between the positions of the harmonics of the pitch modified signal and the LP spectrum of the original signal introduce gain variations, which is more pronounced in the case of female speech [Proc. Int. Conf. on Acoust. Speech and Signal Process. (1997) 1623]. This is minimised by modifying the radii of the poles of the filter to broaden the otherwise peaky linear predictive spectrum. The modified LP coefficients are used for both inverse and forward filtering. This pitch modification scheme is used in our Concatenative Speech synthesis system for Kannada. The technique has also been successfully applied to creating interrogative sentences from affirmative sentences. The modified speech has been evaluated in terms of intelligibility, distortion and speaker identity. Results indicate that our scheme results in acceptable speech in terms of all these parameters for pitch change factors required for our speech synthesis work.

Reconstruction of speech from whispers

This paper investigates a method for the real-time reconstruction of normal speech from whispers. This system could be used by aphonic individuals as a voice prosthesis. It could also provide improved verbal communication when normal speech is not appropriate. The normal speech is synthesized using the mixed excitation linear prediction model. Differences between whispered and phonated speech are discussed and methods for estimating the parameters of this model from whispered speech for real-time synthesis are proposed. This includes smoothing the noisy linear prediction spectra, modifying the formants, and synthesizing of the excitation signal. Trade-offs between computational complexity, delay, and accuracy of different methods are discussed.

Modification of formants in the line spectrum domain

A method for modifying formant locations and bandwidths directly in the line spectrum domain is developed. This method is based on first-order approximations between different representations of the linear prediction spectrum and does not require finding the roots of the prediction polynomial. In addition, this method is less sensitive to pole interaction problems associated with direct pole modification.

Order determination and optimum harmonic reconstruction of quasi-periodic signals in noise

This article proposes a new method for determining the order of wide-band quasi-periodic signals from frequency estimates provided either by their short-time Fourier or linear prediction (LP) spectra. The method consists in the search for harmonic patterns in the signal spectrum that minimize an error sum of the estimated frequencies. This error can be thought of as an extension to the greatest common divisor of a set of spectral lines and operationally resembles to a spectral comb with teeth aperture and width varying as a function of frequency and comb order. The method performs well in noisy environments and when LP spectra are used the method is insensitive to LP model order selection. After the best harmonic structure has been estimated the original signal can be reconstructed with a minimum number of parameters.

SVD based technique for noise reduction in electroencephalographic signals

Noise reduction is a subject of importance in many applications of signal processing such as communication, instrumentation, etc. In a recent paper (Callaerts et al., 1988), singular value decomposition (SVD) has been shown to be a very robust and numerically reliable tool for extracting the desired signals from noisy data. Along similar lines, in this paper, we explore the application of SVD for extracting the actual electroencephalographic (EEG) signals from electrooculogram (EOG) contaminated EEG signals. The key point which made the SVD technique suitable for this EEG enhancement problem is the orthogonality between the EEG and EOG signals. The problem has been set up in such a way that the SVD technique could be applied directly to the recorded signals. Studies were performed on simulated as well as real signals. Signal to noise ratio (SNR) and linear prediction (LP) spectra (along with time plots) are used as measures for comparing the performance of the proposed algorithm for the cases of one EOG channel and two EOG channels. Studies on simulated as well as actual signals show that the technique is very effective in minimizing EOG artefacts from noisy EEG signals.

ANC Schemes for the Enhancement of EEG Signals in the Presence of EOG Artifacts

One of the most important applications of adaptive systems is in noise cancellation using adaptive filters. In this paper, we propose adaptive noise cancellation schemes for the enhancement of EEG signals in the presence of EOG artifacts. The effect of two reference inputs is studied on simulated as well as recorded EEG signals and it is found that one reference input is enough to get sufficient minimization of EOG artifacts. This has been verified through correlation analysis also. We use signal to noise ratio and linear prediction spectra, along with time plots, for comparing the performance of the proposed schemes for minimizing EOG artifacts from contaminated EEG signals. Results show that the proposed schemes are very effective (especially the one which employs Newton's method) in minimizing the EOG artifacts from contaminated EEG signals.

A non-linear estimation model for adaptive minimization of EOG artefacts from EEG signals

In this paper, we propose an adaptive noise cancellation scheme in a novel way for the minimization of electrooculogram (EOG) artefacts from corrupted EEG signals. This method is based on the fact that the transfer function of the biological neuron can be modeled as a sigmoid non-linearity. Comparison of the time plots and the smoothed linear prediction spectra show that the proposed method effectively minimizes the EOG artefacts from corrupted EEG signals. We have also studied the performance of the above scheme for different values of filter order ( P) and the convergence factor (μ). Normalised Mean Squared Error (NMSE) has been used as the measure for comparison. The study shows that the NMSE decreases with increase in P and μ (but saturates after certain values of the parameters), thereby implying a better EOG minimization from EEG signals. It is also observed that the EOG minimization scheme with two EOG reference inputs works better than that with one reference input.

Flexible vowel recognition by the generation of dynamic coherence in oscillator neural networks: speaker-independent vowel recognition.

We propose a new model for speaker-independent vowel recognition which uses the flexibility of the dynamic linking that results from the synchronization of oscillating neural units. The system consists of an input layer and three neural layers, which are referred to as the A-, B- and C-centers. The input signals are a time series of linear prediction (LPC) spectrum envelopes of auditory signals. At each time-window within the series, the A-center receives input signals and extracts local peaks of the spectrum envelope, i.e., formants, and encodes them into local groups of independent oscillations. Speaker-independent vowel characteristics are embedded as a connection matrix in the B-center according to statistical data of Japanese vowels. The associative interaction in the B-center and reciprocal interaction between the A- and B-centers selectively activate a vowel as a global synchronized pattern over two centers. The C-center evaluates the synchronized activities among the three formant regions to give the selective output of the category among the five Japanese vowels. Thus, a flexible ability of dynamical linking among features is achieved over the three centers. The capability in the present system was investigated for speaker-independent recognition of Japanese vowels. The system demonstrated a remarkable ability for the recognition of vowels very similar to that of human listeners, including misleading vowels. In addition, it showed stable recognition for unsteady input signals and robustness against background noise. The optimum condition of the frequency of oscillation is discussed in comparison with stimulus-dependent synchronizations observed in neurophysiological experiments of the cortex.