Line Spectrum Pair Research Articles

Transparent quantization of speech LSP parameters based on KLT and 2-D-prediction

Abstruct- In this correspondence, a two-stage approach based on Karhunen-Loeve transform and 2-D prediction is proposed for efficient quantization of line spectrum pair (LSP) parameters of speech. Besides, a switched classifier is incorporated with this approach to reduce the outlier frames (spectral distortion greater than 2 dB) down to about 0.27% and to eliminate frames with spectral distortion greater than 4 dB at an average bit-rate below 19 b/frame.

Equivalent distortion measures for quantisation of LPC model

The authors show that for small errors in an LPC model, spectral distortion and weighted squared distortion measures in the coefficients of arbitrary one-to-one transformations of the LPC parameters are equivalent. For line spectrum pairs and immitance spectrum pairs, the weighting matrix is diagonal, which is advantageous for efficient quantisation.

Efficient scalar quantisation of line spectrum pairs using the ordering property

An efficient quantisation method of line spectrum pairs (LSP) which has good performance and very low complexity and memory is proposed. The ordering property of the LSP parameters is utilised in the DPCM scheme. The new scalar quantisation algorithm requires 32 bit/frame to achieve 1 dB2 average spectral distortion. The quantisation performance has also been shown to be robust across databases and different speakers.

Dynamic partial search scheme for stochastic codebook of FS1016 CELP coder

The authors present a new classified dynamic partial search structure for the stochastic codebook of the FS1016 CELP coder to replace the fixed partial search for selecting the best excitation vector of the stochastic codebook. In the proposed scheme, the conventional one-stage stochastic codebook search is substituted with a two-stage dynamic method for reducing the computational complexity without degrading the voice quality. The establishment of this structure is based on two classifiers, one for the line spectrum pairs (LSP) of the input signals, and the other for the autocorrelation coefficients (AC) of the stochastic codebook search target. In addition, the stochastic codebook is classified into K subcodebooks, and with these two classifiers it is possible to determine dynamically which subcodebook needs to be searched. This method achieves a reduction in the search procedure by a factor of 2-8. The efficiency of these two classifiers is discussed and the comparison of the performance between the fixed partial search and the proposed technique is also addressed.

Optimal transform coding for speech line spectrum pair parameters based on spectral-weighted error criterion

The line spectrum pair (LSP) is one of the most popular and efficient parameters for representing the short-time spectrum of speech signal. About 34 bits/frame is needed for direct scalar quantization of LSP parameters to maintain a good quality. Based on the spectral-weighted Euclidean distance of LSP parameters, a modified transform coding of LSP parameters which takes advantage of both strong interframe and intraframe correlations is proposed in this paper. This method incorporated with the DPCM scheme does not introduce further buffering delay. It utilizes the spectral-weighted LSP parameters in transform coding instead of LSP vector itself and needs only to quantize the prediction residuals in transformed domain. The much smaller dynamic ranges and symmetrically bell-shaped distributions of those residuals explain the success of this scheme. With the frame period of 10 ms, the spectral distortion limen of 1 dB2 can be reached at 18 bits/frame with inside training test data and 19 bits/frame with outside training test data.

Enumeration and trellis-searched coding schemes for speech LSP parameters

Speech coders employing forward adaptive predictive coding (APC) and operating at medium-to-low bit rates necessitate efficient encoding of the linear predictive coding (LPC) coefficients. Line spectrum pair (LSP) parameters are currently one of the most efficient choices of transmission parameters for the LPC coefficients. The authors briefly reviews LSP parameters and presents several low-delay coding schemes for the parameters. The coders are simulated using data generated from both the autocorrelation and covariance LPC analysis methods. The performances of the coders are given for a variety of rates and LPC analysis conditions. The most efficient scheme developed uses a uses a predictive form of trellis-coded quantization (TCQ). Its performance is comparable or superior to that of other low-delay LSP coding schemes. An enumeration scheme that reduces the rate of a given scalar quantization structure without decreasing coder performance is also presented. >

Pattern matching vocoder using LSP parameters

The system utilizes a linear predictive coding (LPC) analyzer, an Attenuator, a line spectrum pair (LSP) analyzer, a reference pattern memory and a pattern matching device. The LPC analyzer derives LPC parameters from an input speech signal. The LPC parameters are attenuated in the attenuator and fed to the LSP analyzer for deriving LSP parameters which are in turn fed to the pattern matching device. The reference pattern memory stores a plurality of reference patterns composed of a sequence of LSP parameters for a variety of predetermined speech samples. The pattern matching device is connected to the LSP analyzer and the reference pattern memory to select the reference pattern which most closely resembles the input pattern from the LSP analyzer and to provide a label code as an output thereof. On the decoding side, a decoder is responsive to the label for generating LPC parameters corresponding to the reference pattern of the label. A residual signal which is also transmitted with the reference label is received and fed with the generated LPC parameters to a synthesis filter for providing a synthesized speech signal which is subsequently converted into an analog signal.

New LSP encoding method based on two-dimensional linear prediction

A new line spectrum pair (LSP) encoding method, two-dimensional differential LSP (2DdLSP) coding, is proposed. This 2DdLSP approach simultaneously takes advantage of strong interframe and intraframe correlation of LSP parameters. The parameters to be quantised are the prediction residuals which possess the property of smaller variance than LSP parameters. This is the key point for the success of the 2DdLSP method. Three quantisation schemes, a scalar quantisation and two vector quantisations (VQs), are presented. The best result in simulation is achieved by using the product codebook VQ. A spectral distortion of 1 dB2 at 19 bits per frame can be obtained when the frame period is 10 ms. Both within-training and out-of-training tests have demonstrated the robustness of the proposed method to data variation.

Enhanced distance measure for LSP-based speech recognition

A distance measure based on the line spectrum pair (LSP) representation is proposed for improving the performance of speech recognition systems. The quefrency weighted cepstrum is to the root-power sum (RPS) of the inverse filter what the proposed measure is to the RPS of the filter whose roots are on the unit circle and expressed as LSP frequencies. The recognition results on a set of 10 confusable syllables show that the performance of this measure is better than that of the Euclidean distance measure for the LSP.

Optimal quantization of LSP parameters

Two nonuniform aspects of the line spectrum pair (LSP) linear predictive coding (LPC) parameters are investigated, including nonuniform statistical distributions and spectral sensitivities of adjacent LSP frequency differences. Based upon these two nonuniform properties, a globally optimal scalar quantizer is designed for each differential LSP frequency. The design algorithm is dynamic programming based and minimization of a nontrivial data dependent spectral distortion is adopted as the optimality criterion. At 32 bits/frame, the new LSP quantizer achieves a 1-dB average log spectral distortion, a commonly accepted level for reproducing perceptually transparent spectral information. The quantization performance has also been shown to be robust across different speakers and databases.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Fading bit error protection for digital cellular multi-pulse speech coder

Protection of a digital multi-pulse speech coder from fading pattern bit errors common in a digital mobile radio channel is accomplished with error detection techniques which are simple to implement and require no error correcting codes. A synthetic regeneration algorithm is employed which uses only the perceptually significant bits in the transmitted frame. Separate parity checksums for line spectrum pair frequency data, pitch lag data and pulse amplitude data are added to each frame of speech coder bits in the transmitter. The bits are then transmitted through a mobile environment susceptible to fading that induces bursty error patterns in the stream. At the receiving station, the parity checksum bits and speech coder bits are used to determine if an error has occurred in a particular section of the bit stream. Detected errors are flagged and supplied to the speech decoder. The speech decoder uses the error flags to modify its output signal so as to minimize perceptual artifacts in the output speech. Separate checksums are developed for subsets of line spectrum pair (LSP) coefficients and related speech data, whereby a single subset may be error-detected and replaced, rather than an entire frame.

A new efficient algorithm to compute the LSP parameters for speech coding

In this paper, the split Levinson algorithm is used to develop an efficient algorithm to compute the Line Spectrum Pairs (LSP) in Linear Predictive Coding (LPC) of speech. We propose two new real functions defined from the reciprocal and antireciprocal parts of the predictor polynomials obtained from the split Levinson algorithm. These functions are shown to obey three-term recurrence relations. Thus the LSP parameters are directly available from the eigenvalues of tridiagonal matrices, the entries of which are computed from only one version of the split Levinson algorithm. When compared with other existing methods, this algorithm is better in terms of complexity. Der Split-Levinson-Algorithmus wird benutzt zur Entwicklung eines effizienten Algorithmus' für die Berechnung der Line-Spektrum-Pairs (LSP) bei der linear-prädiktiven Kodierung (LPC) von Sprache. Wir schlagen zwei neue reelle Funktionen vor, die wir mit Hilfe des reziproken und des anti-reziproken Teils der Prädiktor-Polynome definieren, welche man aus dem Split-Levinson-Algorithmus erhält. Es wird gezeigt, daβ diese Funktionen Rekursionsbeziehungen in drei Termen Gehorchen. Auf diese Weise Kann man die LSP-Parameter unmittelbar aus den Eigenwerten tridiagonaler Matrizen erhalten, deren Elemente sich aus nur einer Version des Split-Levinson-Algorithmus berechnen lassen. Im vergleich zu anderen bekannten Methoden weist dieser Algorithmus eine geringere Komplexität auf. Dans ce papier, l'algorithme de Levinson éclaté est utilisé pour développer un algorithme efficace pour le calcul des paires de raies spectrales (LSP) en prédiction linéaire pour le codage de la parole. Nous proposons deux nouvelles fonctions réelles, ayant des propriétés intéréssantes, définies à partir des polynômes préditeurs symétriques et anti-symétriques. Les relations de récurrences à trois termes pour les fonctions réelles sont déduites. Les paramètres LSP sont ensuite calculés directement à partir des valeurs propres de matrices tridiagonales symétriques dont les éléments sont calculés par une seule version de l'algorithme de Levinson éclaté. Comparé à d'autres méthodes existantes, cet algorithme est le moins complexe en terme de nombre total d'opérations.

Joint source and channel trellis coding of line spectrum pair parameters

An innovative method of coding Line Spectrum Pair (LSP) parameters for transmission over noisy channels is presented. Typically, low bit-rate speech coders use these parameters to convey perceptually important spectral information. Thus it is necessary that these parameters are not only efficiently quantized but preserved during transmission. The scheme uses a joint source and channel coding technique applied to a concatenated trellis structure. Operating as a source coder, the system encodes below the 1 dB spectral distortion limit. It is shown that by modifying the encoder cost function to include the expected channel distortion, the code exhibits improved robustness to channel errors. This is accomplished with minimal increase in complexity and without increase in bit-rate. It is noted that the scheme performs well over a wide range of channel bit error rates and compares favourably with a standard scalar LSP quantizer tandemed with a channel coder both in terms of bit-rate and channel noise immunity.

A phase approximation method and stability conditions of all‐pole systems in terms of 2‐D line spectrum pair

AbstractThis paper discusses the stability of a two‐dimensional (2‐D) all‐pole system, especially its relation to the phase characteristic. The discussion starts with Ekstrom's stability condition for the asymmetrical half‐plane model, which is the most general as the 2‐D infinite impulse response (IIR) system. The stability theorem known for the phase characteristic of the 1‐D all‐pole system is extended and applied to the 2‐D case. The 2‐D line spectrum pair is introduced as the intersection of the phase characteristic of the 2‐D all‐pole system and a particular set of planes. The stability theorem for the all‐pole system is derived as the condition for such a system. Then, based on the stability theorem for the 2‐D all‐pole system, an approximation method for the phase by the all‐pole system which ensures the stability is proposed.It is shown that the proposed phase approximation method corresponds to a particular case of the signal restoration using only the phase data given by Oppen‐heim et al. As an improvement of the proposed phase approximation method, the evaluation function by Oppen‐heim based on the error of the equation is replaced by the evaluation function which is the phase error itself. The error minimization method is proposed.Finally, as an application example, the phase compensation problem by the all‐pass system is considered, and the usefulness of the proposed method is demonstrated.

Spectral coding by LSP frequencies—scalar and segmented VQ-methods

Line spectrum pairs (LSP) is one interesting way of representing the filter parameters of a linear production model. In particular, LSPs have been used for quantisation, i.e. for the spectral coding problem of speech coding. The concept of close-to-optimal search strategies, as well as optimal quantiser design, is discussed for the case of LSP-based quantisation. Previous scalar approaches are generalised to a vectorial treatment by carefully selecting a VQ technique that can provide robust results with low complexity. Simulation results are presented both for the scalar and for the vectorial case, showing that a two-dimensional VQ approach can provide substantial bit savings in realistic cases.

A new objective speech quality measure for speech coding systems

This paper presents a new objective measure for speech quality testing to be used with speech coding systems. This measure is presented in the context of an ongoing 15-year study of objective measures for speech quality. As part of this research, well over 3000 measures have been evaluated in a study that has done well over a million separate correlation analyses. Included in the study were parametric forms of virtually all commonly used objective measures as well as a large number of new, complex composite measures. The new objective speech quality measure to be presented is a composite measure based on the line spectrum pair (LSP) decomposition of the linear predictive coding (LPC) polynomial. The measure is intended specifically to be used for the design of LPC quantizers for use in a wide variety of speech coding systems. This new measure is the first measure developed in this research that has both very high performance in terms of its ability to predict listener subjective responses as well as the properties necessary to be used explicitly in the design of LPC quantizers.

A study of line spectrum pair frequencies for vowel recognition

The line spectrum pair (LSP) frequency representation has recently been proposed as an alternative linear prediction (LP) parametric representation. In the context of speech coding, this representation shows better quantization properties than the other LP parametric representations. In the present paper, the LSP representation is studied for speech recognition. Several distance measures based on this representation are investigated on a steady-state vowel recognition task. The weighted LSP distance measure is found to result in the best performance. The performance of the weighted LSP distance measure is compared with that of the other popular LP distance measures (such as the Itakura, cepstral, weighted cepstral, root-power-sum, log area ratio and reflection coefficient distance measures). The weighted LSP distance measure is found to perform significantly better than these popular LP distance measures.

A perception‐based LSP distance measure for speech recognition

Because of their high correlation and reduced sensitivity to quantization errors, the line spectrum pair (LSP) frequency parameters have been used recently for efficient quantization of LPC information for speech coding. In the present paper, the LSP representation is used for speech recognition and a new perception‐based LSP distance measure is proposed. This distance measure exploits the following two properties of the speech perception process [D. H. Klatt, Proc. ICASSP, 1278–1281 (1982)]: (1) The formant frequencies are the most important parameters for speech perception: and (2) the formant bandwidths and spectral tilt contribute very little to speech perception. The present distance measure uses these two properties in the form of weights in a weighted Euclidean distance measure. In order to derive these weights, the LPC power spectrum P(f) is computed for each speech frame and the weight for a given LSP frequency ft is taken to be proportional to [P(ft)]c. The perception‐based LSP distance measure is studied here on a speaker‐dependent speech recognition task. The optimum value of the exponent c is found here to be 0.15. The perception‐based LSP distance measure results in a recognition score of 95.7%, while the recognition accuracy is found to be 89.9% by using the conventional Euclidean distance measure on LSP parameters and 94.1% by using the liftered cepstral distance measure [B. H. Juang et al., IEEE Trans. Acoust. Speech Signal Process. ASSP‐35, 947–954 (1987)].

Experimentation with synthesized speech generated from line-spectrum pairs

Line-spectrum pairs (LSP's) are frequency-domain parameters similar to formant frequencies. Thus, they have frequency-selective spectral-error characteristics which allow LSP quantization in accordance with auditory perception. In addition, ease of estimating the spectral-error sensitivity of each line spectrum makes possible encoding each line spectrum efficiently. This correspondence, for the first time, demonstrates that a 31 bit representation of LSP's provides similar intelligibility as a 41 bit representation of reflection coefficients in a current 2400 bit/s LPC. Even with a 12 bit quantization of LSP's, the loss of speech intelligibility is minor, only 2.4 points below that of a 41 bit quantization of reflection coefficients as measured by the diagnostic rhyme test (DRT) which tests initial-consonant discrimination.