Pitch Filter Research Articles

보코더에서 프레임별 에너지 보상에 의한 피치검색 성능 개선에 관한 연구

보코더 과정 중 피치를 분석하는 과정은 코드북 분석과 더불어 부호화기 성능에 큰 영향을 미치는 부분이다. 피치를 검색하기 위해 먼저 신호의 주기성을 강조해야 하는데 일반적으로 자기상관법의 일종인 듀얼 펄스 알고리즘 기법을 이용하여 주기성을 강조한 후 피치를 검색하는 방법을 많이 사용한다. 이 방법은 한 프레임 내에서 두 개의 펄스 간격을 변경시키면서 음성신호와의 상관관계 값을 구하여 상관관계가 가장 높을 때의 펄스 간격을 찾는다. 또한 상관관계 값이 가장 높을 때의 주기가 가장 뚜렷한 반복 구간인데 이것을 피치 주기라 한다. 이때 반주기, 배주기 및 세배주기가 주된 주기로 찾아지는 경우에는 이 간격을 피치 주기라 할 수 없어 이를 해결하기 위한 여러 가지방법들이 제안되어 있다. 본 논문에서는 피치를 검색하기 전에 프레임내의 전체 에너지변화 비율을 추정하여 신호의 에너지 레벨을 미리 보상해 준 후 피치를 검색하는 방법을 제안한다. 이 방법을 적용하면 피치검색 시간을 단축할 수 있고, 피치 검색의 정확도를 높일 수 있어 전반적인 피치 검색에 관한 성능이 개선된다. It is important to search a pitch for vocoder. The major drawback to vocoders is their large computational requirements in searching a pitch and a codebook. In this paper, a simple method is proposed to improve the pitch searching process in the pitch filter almost without degradation of quality. The period of speech signal is emphasized by using Dual Pulse technique, the same type of autocorrelation method, in pitch search. Sometimes the incorrect pitch can be obtained by halving, doubling and trifling, To solve it, before searching a pitch, we estimate energy rate in a frame and compensate envelop of signal with it. By using the proposed algorithm in pitch search, its required computation are reduced and searching pitch is improved.

Voice synthesis system utilizing a transfer function

A voice synthesis system is fundamentally configured by a sound-source model, which simulates human voices and the like, and a voice-path model which simulates properties of voice paths between vocal cords and lips. The sound-source model is embodied by a code book which stores a plurality of code words, representative of waveform patterns, with respect to each of the voices. Each of the code words is selected by an information index. The voice-path model is embodied by a full-pole synthesis filter whose characteristic curve provides multiple poles, each of which is represented by polar coordinates. There is further provided a pitch filter and an all-pass filter. Data representative of the code word selected is supplied to the pitch filter, in which a first delay time, set by a number of delay-time units, is imparted to the data. Then, the all-pass filter imparts a second delay time, which is smaller than the delay-time unit, to the data in response to pitch-variation information. Those filters are provided to perform a fine adjustment of the pitch of the data. Thereafter, the full-pole synthesis filter performs filtering processing on the data in accordance with a coefficient which is set in response to the polar coordinates and pitch-variation information. Thus, signals indicative of synthesized sounds are produced by the full-pole synthesis filter.

Efficient pitch filter encoding for variable rate speech processing

Analysis-by-synthesis techniques are used in a wide variety of speech coding standards and applications for rates below 16 kbps. The presence of a long-term predictor, commonly known as the adaptive codebook, is critical to coder performance at the lower rates. Unfortunately, the encoding rate and computational requirements for high-quality encoding of pitch filter parameters can be excessive. Several popular approaches explore the trade-off between predictor order, allocated bit rate, and computational requirements for long-term predictor optimization. We investigate the relative performance of several long-term predictor structures and present a new approach to vector quantization of the pitch filter coefficients having a subjective quality equivalent to other schemes, but at a lower coding rate and requiring significantly less closed-loop computation. The performance is evaluated in a variable-rate CELP coder at an average rate of 2 kbps and in Federal Standard 1016 CELP.

Method for processing speech signal in speech processing system

A method for processing an input speech signal to be applied to a CELP vocoder has the steps of obtaining preliminary pitch search intervals by a preprocessing autocorrelation expression from a pitch lag of a synthesized speech signal which is synthesized from a residual signal of the input speech signal; computing coefficients of pitch filter with respect to the preliminary pitches; searching a high interval in the autocorrelation; and removing the remaining interval other than the high interval in the pitch lag. Since the present invention proposes a speech processing method which uses only a high interval in autocorrelation of a voice waveform in pitch-searching, and where such a speech processing method is embodied in a CELP vocoder, total computation time of the CELP vocoder can be decreased 37% or more without lowering speech quality. Therefore, a digital signal processor, which is low in price and is slow in speed, can be embodied in a CELP vocoder.

CS-ACELP speech compression system with adaptive pitch prediction filter gain based on a measure of periodicity

A speech coding system employing an adaptive codebook model of periodicity is augmented with a pitch-predictive filter (PPF). This PPF has a delay equal to the integer component of the pitch-period and a gain which is adaptive based on a measure of periodicity of the speech signal. In accordance with an embodiment of the present invention, speech processing systems which include a first portion comprising an adaptive codebook and corresponding adaptive codebook amplifier and a second portion comprising a fixed codebook coupled to a pitch filter, are adapted to delay the adaptive codebook gain; determine the pitch filter gain based on the delayed adaptive codebook gain, and amplify samples of a signal in the pitch filter based on said determined pitch filter gain. The adaptive codebook gain is delayed for one subframe. The pitch filter gain equals the delayed. adaptive codebook gain, except when the adaptive codebook gain is either less than 0.2 or greater than 0.8., in which cases the pitch filter gain is set equal to 0.2 or 0.8, respectively.

Voice compression system and method

A system for performing voice compression which includes, a conversion means for converting analog voice signals into discrete samples of digital voice data and collecting the discrete samples into segments, with a means for dividing the segments into subsegments and for producing therefrom a current voice subsegment. A pitch prediction means is used for determining the long term predicted gain of the current voice subsegment by comparing the current voice subsegment to reconstructed voice samples to produce a pitch predictor gain and a lag component and a pitch filter means is used for pitch filtering the current voice subsegment based on the pitch predictor gain and the lag component which then produces long term residual samples. The system includes a means for scaling the long term residual samples based on the peak amplitude of the long term residual samples to produce normalized long term residual samples and a means including a code book stored in a memory for locating a single memory address of a closest match between the normalized long term residual samples and the stored distinct normalized long term residual samples. The system includes an output means for providing the distinct memory address, the pitch predictor gain, the lag component and the peak amplitude for each voice subsegment and a means to reconstruct these samples based on these parameters.

Analysis-by-synthesis 2,4 kbps linear predictive speech codec

First Page

Pseudo-multi-tap pitch filters in a low bit-rate CELP speech coder

The pitch filter in a low bit-rate CELP speech coder has a strong impact on the quality of the reconstructed speech. In this paper we propose a pseudo-multi-tap pitch filter with fewer degrees of freedom than the number of prediction coefficients, but which gives a higher pitch prediction gain and a more appropriate frequency response than a conventional one-tap pitch filter. First, we present an analysis model for the pseudo-multi-tap pitch prediction filter. Then, we introduce a pseudo-multi-tap pitch prediction filter with a fractional pitch lag. The prediction gain of the pseudo-multi-tap pitch filter is compared to that of conventional one-tap and three-tap pitch filters with integer and non-integer pitch lags. A switching configuration is also studied. This filter switches modes depending on the prediction gain. The stability of a pseudo-multi-tap pitch synthesis filter in a CELP coder is considered. We propose a stabilization method with a relaxed stability test. This relaxed test gives better results than a strict stability test. Finally, we have incorporated the pseudo-multi-tap pitch filter into a 4.8 kbit/s CELP speech coder. Both the objective SNR and subjective quality are better than for a conventional one-tap pitch filter.

On encoding pitch and LPC parameters for low‐rate speech coders

AbstractThis paper describes the development of efficient methods to encode the parameters of long‐term (pitch) and short‐term (LPC all‐pole) prediction filters with a limited bit‐rate, such that the design of high quality speech coders in the 3–4 kbit/s range becomes easier. Our results indicate that good performance can be reached with about 800 bit/s to represent both the pitch lags and the coefficients of a three‐tap pitch filter, thus achieving a significant 36% bit‐rate reduction in comparison with conventional quantization techniques, and using 24 bit/frame to encode spectral parameters relative to a multilanguage speech database.

Fast methods for code search in CELP

The code excited linear prediction (CELP) technique has the potential for producing high quality synthetic speech at bit rates as low as 4.8 kb/s. Three fast search methods for selecting an optimal excitation sequence from a code book of stochastic vectors are described. The key idea is to inverse-filter the actual speech with the formant and pitch filters to produce a residual error sequence (RES). The residual error is used to identify a neighborhood or a subset of codes for further processing. The first method, called dynamic nearest neighborhood (DNN), attempts to dynamically construct a neighborhood of the K codes of maximum correlation with the residual error. The second method, called nearest fixed neighborhood (NFN), clusters the code book into a fixed number of cells, and code search is performed on the codes of the cell nearest to the RES. The two methods achieve a reduction in the search procedure by a factor of 8-20 times. The third method combines the advantages of the first two methods to attain a reduction from 40 to 50 times. The performance of these techniques and some of their ramifications are addressed. >

A low delay 16 kb/s speech coder

A code tree generated by a stochastically populated innovations tree with a backward adaptive gain and backward adaptive synthesis filters is considered. The synthesis configuration uses a cascade of two all-pole filters: a pitch (long time delay) filter followed by a formant (short time delay) filter. Both filters are updated using backward adaptation. The formant predictor is updated using an adaptive lattice algorithm. The multipath (M, L) search algorithm is used to encode the speech. A frequency-weighted error measure is used to reduce the perceptual loudness of the quantization noise. The addition of the pitch filter gives 2-10-dB increase in segSNR (segmental signal-to-noise ratio) in the voiced segments. Subjective testing has shown that the coder attains a subjective quality equivalent to 7 b/sample log-PCM (pulse code modulation) with an encoding delay of eight samples (1 ms with an 8-kHz sampling rate). >

Joint optimization of linear predictors in speech

Formant and pitch predictors which are jointly optimized are discussed. The first configuration considered is a combination prediction error filter (in either a transversal or lattice form) that performs the function of both a formant and a pitch filter. Although a transversal combination filter outperforms the conventional F-P (formant followed by pitch) sequential solution, the combination filter exhibits a high incidence of nonminimum phase filters. For an F-P cascade connection, combined solutions and iterated sequential solutions are found. They yield high prediction gains than the conventional F-P sequential solution. Furthermore, a practical implementation of the iterated sequential solution is developed such that both the formant and pitch filters are minimum phase. This implementation leads to decoded speech of higher perceptual quality than the conventional sequential solution.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Pitch prediction filters in speech coding

Prediction error filters which combine short-time prediction (formant prediction) with long-time prediction (pitch prediction) in a cascade connection are examined. A number of different solution methods (autocorrelation, covariance, Burg) and implementations (transversal and lattice) are considered. It is found that the F-P cascade (formant filter before the pitch filter) outperforms the P-F cascade for the both transversal- and lattice-structured predictors. The performances of the transversal and lattice forms are similar. The solution method that yields a transversal structure requires a stability test and, if necessary, a consequent stabilization. The two cascade forms are implemented as part of an APC coder to evaluate their relative subjective performance. >

Stability and performance analysis of pitch filters in speech coders

This paper analyzes the stability and performance of pitch filters in speech coding when pitch prediction is combined with formant prediction. A computationally simple stability test based on a sufficient condition is formulated for pitch synthesis filters. For typical orders of pitch filters, this sufficient test is very tight. Based on the test, a simple stabilization technique that minimizes the loss in prediction gain of the pitch predictor is employed to generate stable synthesis filters. Finally, it is observed that the quality of decoded speech improves significantly when stable synthesis filters are employed.

Pitch identification device

A transducer receives notes from a musical instrument and converts them into electrical signals. The signals are passed to filter banks containing filters having passbands in different octave ranges. The filtered signals in each bank are added and translated in frequency to the frequency range of the highest octave of interest. The translated signals are then passed through a plurality of filters each of which is tuned to pass directly the frequency of one pitch in that octave. Indicators are attached to the pitch filters to indicate the presence of a signal passing therethrough. Also, additional indicators may be attached to the octave filters to indicate the octave from which the signal originated.