Linear Prediction Vocoder Research Articles

A NEW DISTORTION MEASURE FOR PARAMETER QUANTIZATION BASED ON MELP

Parameter quantization is very important for the synthetic speech quality of the vocoder. A new distortion measure for pitch as well as lsf quantization in ultra low bit rate Vocoder, whose parameters for several consecutive frames are grouped into a vector and jointly quantized to obtain high coding efficiency, is proposed based on mixed excitation linear prediction(MELP) vocoder. The product of sum of band pass voicing coefficients and gain parameter is used to denote the weighting factor of pitch as well as lsf parameters of current speech frame in the consecutive frames using weighted squared Euclidean distance measure to search the vector codebook. Comparing with the traditional method for a constant weighting factor by distinguishing Voiced/Unvoiced(UV) pattern of each speech frame, objective test results show that the quantization distortion of pitch is reduced by 3.3% and the mean opinion score (MOS) is increased by almost 0.1(3.5%).

DCT_M model for excitation parameter in low bit rate vocoder

The description precision of an excitation signal greatly influences the quality of reconstructed speech in low bit rate vocoders. To improve the reconstruction quality, the DCT_M model is proposed to express the excitation spectral parameter, which transforms the variable length vector to fixed dimension vector through DCT transformation. It then quantizes the fixed length vector using multi-stage vector quantization. Tests show that the proposed method can keep the shape of the entire spectral envelope and reduce model error thus greatly improve the description precision. Test results in the sine excitation linear prediction (SELP) vocoder show that the DCT_M model can improve the naturalness of reconstructed speech, with subjective test score of 65%.

An improved 2.4kb/s Mixed Excitation Linear Prediction vocoder

This letter presents two improvements on 2.4 kb/s Mixed-Excitation Linear Prediction (MELP) vocoder. The one is a new parameter R edzc named energy to differential zero-crossing rate which is used in adaptation of V/UV decision of transitional segments and low energy level speech segments. The other is a multi-path searching method for Multi-Stage Vector Quantization (MSVQ) of line spectral frequency. Subjective tests show that the intelligiblity and naturallity of improved MELP vocoder are preferable to those of the original one.

Channel and source considerations of a bit-rate reduction technique for a possible wireless communications system's performance enhancement

In wireless commercial and military communications systems, where bandwidth is at a premium, robust low-bit-rate speech coders are essential. They operate at fix bit rates and those bit rates cannot be altered without major modifications in the vocoder design. A novel approach to vocoders, in order to reduce the bit rate required to transmit speech signal, is proposed. While traditional low-bit-rate vocoders code original input speech, the proposed procedure operates on the time-scale modified signal. The proposed method offers any bit rate from 2400 b/s to downwards without modifying the principle vocoder structure, which is the new NATO standard, Stanag 4591, Mixed Excitation Linear Prediction (MELP) vocoder. We consider the application of transmitting MELP-encoded speech over noisy communication channels by applying different modulation techniques, after time-scale compression is applied. Three different time-scale modification algorithms have been evaluated and waveform similarity overlap and add (WSOLA) algorithm has been selected for time-scale modification purposes. Computer simulation results, both source and channel, are presented in terms of objective speech quality metrics and informal subjective listening tests. Design parameters such as codec complexity and delay are also investigated. Simulation results lead to a possible wireless communications system, whose performance might be enhanced by using the spared bits offered by the procedure.

Channel and source considerations of a MELP derived vocoder operating at reduced bit rates

Robust low bit rate speech coders are essential in commercial and military communication systems. They operate at fixed bit rates and those bit rates cannot be altered without major modifications in the vocoder design. In this paper we introduce a scaled speech coder, which operates on time-scale modified input speech. The proposed method offers any bit rate from 2400 b/s to downwards without modifying the principle vocoder structure, which is the mixed excitation linear prediction (MELP) vocoder. We consider the application of transmitting MELP-encoded speech over noisy communication channels after time scale compression is applied. Computer simulation results, both source and channel, are presented in terms of objective speech quality metrics and informal subjective listening tests. A statistical tool called bootstrap is also used to determine the accuracy of these test results. Design parameters such as codec complexity and delay are also investigated.

A flexible bit rate switching method for low bit rate vocoders

Robust low bit rate speech coders are essential in commercial and military communication systems. They operate at fixed bit rates and bit rates cannot be altered without major modifications in the vocoder design. In this paper we introduce a novel approach to vocoders by coding the time-scale modified input speech signal. The proposed method offers any bit rate from 2400 to 720 bits/ s without modifying the vocoder structure. Simulation results, which mainly concentrate on intelligibility, talker recognisability and voice quality versus codec complexity and delay, are also presented. The proposed scaled speech coder delivers communication quality speech at half the bit rate of the new US Federal Standard, mixed excitation linear prediction vocoder.

Model of the human speech production based on vocoder techniques

A tutorial is presented with which the student first gets basic knowledge about the human speech production. It is explained that speech is generated by an air flow coming from the lungs which is, in the case of voiced sounds, periodically interrupted by the vocal cords and then modulated with the resonance rooms of the mouth and nasal tract. For voiceless sounds (like ‘‘f,’’ ‘‘sch’’) the vocal cords are open and the air flow excites the mouth and nasal tract with a noise-like pressure signal. It is therefore adequate to distinguish two components in the speech production: the excitation and the articulation component. In our tutorial, the two components are realized (simulated) by a technical system, known as linear predictive vocoder (LPC vocoder). Its principle is explained by a short text and some graphics and the analogy to the human speech organs is clearly shown. The main emphasis is put on the control parameters of the system—in the main the pitch frequency and the vocal tract parameters (prediction coefficients)—and the effect when.

Design of Chip Set for CDMA Mobile Station

In this paper, we present a design of modem and vocoder digital signal processor (DSP) chips for CDMA mobile station. The modem chip integrates CDMA reverse link modulator, CDMA forward link demodulator and Viterbi decoder. This chip contains 89,000 gates and 29 kbit RAMS, and the chip size is 10 mm x 10.1 mm which is fabricated using a 0.8 μm 2 metal CMOS technology. To carry out the system-level simulation, models of the base station modulator, the fading channel, the automatic gain control loop, and the microcontroller were developed and interfaced with a gate-level description of the modem application specific integrated circuit (ASIC). The Modem chip is now successfully working in the real CDMA mobile station on its first fab-out. A new DSP architecture was designed to implement the Qualcomm code exited linear prediction (QCELP) vocoder algorithm in an efficient way. The 16 bit vocoder DSP chip has an architecture which supports direct and immediate addressing modes in one instruction cycle, combined with a RISC-type instruction set. This turns out to be effective for the implementation of vocoder algorithm in terms of performance and power consumption. The implementation of QCELP algorithm in our DSP requires only 28 million instruction per second (MIPS) of computation and 290 mW of power consumption. The DSP chip contains 32,000 gates, 32k (2k x 16 bit) RAM, and 240k (10kx24 bit) ROM. The die size is 8.7 mmx8.3 mm and the chip is fabricated using 0.8 μm CMOS technology.

Quad-band excitation for low bit rate speech coding.

The excitation parameters of a multiband excitation model for speech production consist of a pitch period and a series of voiced/unvoiced (v/uv) decisions for a number of fixed frequency bands around the harmonics of the fundamental. In this paper, it is shown that, at most, four v/uv variable-length nonoverlapping frequency bands spanning over the telephone bandwidth are adequate to represent the excitation spectrum that provides an improved modeling of mixed voicing and noisy speech. A quad-band excitation (QBE) method is presented which generates excitation source for a linear prediction vocoder for low bit rate transmission of speech signal and the DRT scores validate the potential of the method.

Time envelope vocoder, a new LP based coding strategy for use at bit rates of 2.4 kb/s and below

This paper presents a linear prediction (LP) based vocoder employing a novel technique which ensures smooth evolution of the synthetic speech waveform. In this coder, speech waveforms are considered as having a 'time envelope', the shape of which contains important perceptual information. By ensuring that the time envelope of the synthetic speech closely matches that of the original, natural sounding synthetic speech can be produced. Envelope matching may be achieved using a new, low complexity analysis by synthesis loop at the decoder which determines the synthetic excitation energy. The advantage over more traditional linear prediction vocoders is that the amplitude time envelope is preserved in addition to the spectral envelope, allowing the rapid amplitude transitions associated with onsets to be retained in the synthetic speech, resulting in a more intelligible output. Simply controlling the overall energy of the synthetic excitation is not sufficient to accurately control the synthetic speech energy. Small changes in linear prediction or pitch parameters due to quantization, for example, can cause variations in the synthetic speech amplitude, especially from one pitch cycle to the next resulting in noisy synthetic speech. The inclusion of an analysis by synthesis loop at the decoder ensures that the synthetic speech amplitude is independent of variations in the pitch period and LP parameters. This paper presents a complete vocoder scheme using time envelope matching, including details of techniques such as parameter interpolation, excitation pulse shaping and pitch tracking which have proven necessary to produce natural sounding synthetic speech at 2.4 kb/s and below.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Speech coding: a tutorial review

The past decade has witnessed substantial progress towards the application of low-rate speech coders to civilian and military communications as well as computer-related voice applications. Central to this progress has been the development of new speech coders capable of producing high-quality speech at low data rates. Most of these coders incorporate mechanisms to: represent the spectral properties of speech, provide for speech waveform matching, and optimize the coder's performance for the human ear. A number of these coders have already been adopted in national and international cellular telephony standards. The objective of this paper is to provide a tutorial overview of speech coding methodologies with emphasis on those algorithms that are part of the recent low-rate standards for cellular communications. Although the emphasis is on the new low-rate coders, we attempt to provide a comprehensive survey by covering some of the traditional methodologies as well. We feel that this approach will not only point out key references but will also provide valuable background to the beginner. The paper starts with a historical perspective and continues with a brief discussion on the speech properties and performance measures. We then proceed with descriptions of waveform coders, sinusoidal transform coders, linear predictive vocoders, and analysis-by-synthesis linear predictive coders. Finally, we present concluding remarks followed by a discussion of opportunities for future research. >

Multiband linear predictive speech coding at very low bit rates

The authors describe the multiband linear predictive (MB-LPC) vocoder and its operation at 2.4 kb/s and 1.2 kb/s. The MB-LPC vocoder uses mixed excitation and exploits the advantages of both time and frequency domain speech coding techniques to produce natural sounding, good quality speech. Subjective performance of speech at 2.4 kb/s produced by the MB-LPC is very close to that for the 4.15 kb/s INMARSAT-M IMBE speech coder. Informal listening tests have shown that in most cases people could not tell the difference between the new 2.4 kb/s MB-LPC coder and the 4.15 kb/s INMARSAT-M IMBE coder.

Methods for waveform interpolation in speech coding

Most speech coding algorithms operating at rates of around 8 kb/s attempt to reproduce the original speech waveform. Their efficiency of reproducing the waveform is obtained by using models which exploit knowledge of the generation of the speech signal. In contrast, most coders operating at rates of around 2.4 kb/s are completely parametric, usually transmitting parameters describing the pitch and the spectral envelope at regular intervals. However, because of model inadequacies, the quality of reconstruction of current parametric methods never reaches that of the original signal, even at high bit rates. In this paper, a new method which is positioned between the waveform coders and the parametric coders is presented. It is based on the assumption that, for voiced speech, a perceptually accurate speech signal can be reconstructed from a description of the waveform of a single, representative pitch cycle per interval of 20-30 ms. Figure 1 shows the smooth evolution of the shape of the pitch cycle, which is typical for voiced speech signals. We will show how such a signal can be reconstructed by interpolatingprototype pitch cycles between the updates. The prototypewaveform interpolation (PWI) method retains the natural quality typical of coders which encode the entire waveform, but requires a bit rate close to that of the parametric coders. We discuss PWI methods based on linear prediction (LP). In LP-based speech coders, the signal is reconstructed from knowledge of the predictor coefficients and a description of the excitation signal. Of the existing LP-based algorithms, the code-excited linear-prediction (CELP) algorithm [l] and the LP vocoder [ 2 ] are examples of waveform and parametric coders, respectively. In the simplest form of CELP the speech waveform is described by time-varying LP filter coefficients and a filter excitation consisting of the concatenation of scaled fixed-length vectors from a codebook. To achieve high efficiency during voiced speech, most implementations include a long-term predictor [ 3 1, or adaptive codebook [ 41, to facilitate periodicity of the reconstructed signal. Despite recent improvements [ 5,6], inaccurate reproduction of the periodicity remains the main source of perceptual distortion in the current CELP algorithms at rates below 6 kb/s. In the LP-based vocoders the voiced speech signal is modeled by a single pulse per pitch cycle. Because of excessive periodicity, this often leads to a buzzy character of the reconstructed speech. Recent work has shown that the speech quality can be improved significantly by adding more information about the evolving waveform shape. Using a cluster of pulses for each pitch cycle, with blockwise shape adaptation, in combination with a smoothly varying overall gain produced good results [ 71. Alternatively, good-quality voiced speech can be obtained at rates of around 3 kb/s by careful placement of the single-pulse locations [ 8,9]. Although significantly improved over the LP-based vocoders, and similar in quality to 4.8 kb/s CELP, such single-pulse excited ( SPE) speech coders still suffer from some buzziness. Both the CELP and the SPE methods attempt to reproduce the original waveform by using a (spectrally weighted) signal-to-noise ratio (SNR) of the reconstructed speech signal as a criterion to determine the excitation sequence. However, maintaining the periodicity of the original speech signal is important for its perceptual quality, and maximization of the SNR often leads to a nonoptimal degree of periodicity. Thus, it was found in both the CELP [ 61 and the SPE coders [ 91 that improved speech quality can be obtained by increasing the periodicity, despite an associated reduction in SNR.

High quality multiband LPC coding of speech at 2.4 kbit/s

Until recently good quality digital speech transmission was not possible below about 9.6 kbit/s. With the development of code excited linear prediction (CELP) and multiband excitation (MBE) vocoders,2 high quality digital speech transmission became possible at bit rates as low as 4.8 kbit/s. Below 4.8 kbit/s however, vocoders such as LPC-103 and the channel vocoder4 which can only produce synthetic and unnatural speech are still in use. A new coding scheme termed the multiband LPC (MB-LPC) vocoder is proposed, which produces good and natural quality speech at 2.4 kbit/s. Subjective performance of speech at 2.4kbit/s produced by the CELP is very close to MBE coded speech at 4.8 kbit/s. Informal listening tests have shown that in most cases people could not tell the difference between the new 2.4kbit/s MB-LPC coder and the 4.8 kbit/s MBE vocoder.

Voice encoder and synthesizer

A very small, very flexible, high-quality, linear predictive vocoder has been implemented with commercially available integrated circuit. This fully digital realization is based on a distributed signal processing architecture employing three commercial Signal Processing Interface (SPI) single chip microcomputers. One SPI implements a linear predictive speech analyzer (18), a second implements a pitch analyzer (20), while the third implements the excitation generator and synthesizer (28).

Error correction/masking for digital voice transmission over the land mobile satellite system

The development of a 2400-b/s speech digitizer which provides an acceptable level of intelligibility and quality over land mobile satellite channels is described. Performance tests over simulated channels in the UHF band (800 MHz) are presented. The voice digitizer is a linear prediction (LPC) vocoder which uses a channel error correction and concealment procedure tailored to error statistics for a minimum-shift keyed (MSK) downlink to a moving vehicle. The error-handling technique is based on perceptual criteria and utilizes the parametric nature of LPC representation of speech. A single-error-correcting, single-burst-detecting (28, 20) fire code is shown to be the best choice for the application. The intelligibility of the vocoder is measured and compared to the standard LPC-10 algorithm. The major remaining sources of speech quality degradation due to channel errors are determined and ranked.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

An investigation of mixed‐source excitation models for linear predictive vocoders

Two methods were investigated for modeling the excitation for linear predictive vocoders as a mixture of pulses (periodic component) and noise (nonperiodic component). In one method, linear filters were used to separate the periodic and nonperiodic components of the residual signal. In another method, the two components were separated by locating the areas of peak energy in the residual and labeling the portion of the residual within each pitch period peak as the periodic component and the remainder of the signal as the nonperiodic component. Thus, in the first method, a frequency domain separation of the periodic and nonperiodic component was assumed whereas, in the second method, a time‐domain separation of the two components was assumed. The time‐domain method appeared to be more feasible. Speech was synthesized using a number of parameters to represent the two components of the excitation. Although some improvements in speech quality could be obtained by using very detailed representations of the pulses, less‐detailed representations resulted in poorer quality speech than that obtained using a standard impulse excitation.

Speech quality evaluation using "phoneme-specific" sentences.

A new approach is described for the design of speech materials used in subjective speech quality evaluation. Speech sounds are classified by their acoustic properties, and sentences are composed so as to concentrate all sounds with similar properties within one sentence. As a test of the method, subjective quality data were collected, using both a rank ordering and a rating task, from a set of 12 linear predictive vocoders, whose parameters were chosen so as to equate their bit rates at 2600 bps. The results show that the method can reliably reveal small differences in quality, and also yields information that can be of diagnostic help in determining the causes of quality degradation by a particular vocoder. A set of phoneme-specific sentences is appended.

Improvement of the excitation source in the narrow-band linear prediction vocoder

The major weakness of the current narrow-band LPC synthesizer lies in the use of a canned invariant excitation signal, The use of such an excitation signal is based on three primary assumptions, namely, 1) that the amplitude spectrum of the excitation signal is flat and time invariant, 2) that the phase spectrum of the voiced excitation signal is a time-invariant function of frequency, and 3) that the probability density function of the phase spectrum of the unvoiced excitation signal is also time invariant. This paper critically examines these assumptions and presents modifications which improve the quality of the synthesized speech without requiring the transmission of additional data. Diagnostic acceptability measure (DAM) tests show an increase of up to five points in overall speech quality with the implementation of each of these improvements. These modifications can also improve the speech quality of LPC-based speech synthesizers.

Dynamic data compression algorithms in speech coding

Speech coding is to use a series of bits to encode speech signals, such that it is much easier and more efficient to digitally transmit or store speech signals. Bit rate is the number of bits required for every second of speech signal. Based upon parametric descriptions of speech signals, the linear prediction analysis provides a successful method for low bit rate speech coding. Due to the high redundancy among the parameters of adjacent speech signal frames, further data compression becomes possible. In this paper, various data compression techniques and algorithms are studied and two sets of variable frame rate linear predictive vocoders are developed based on frame interpolation and frame repeating concepts, respectively. Simulation results indicate that almost identical speech quality can be achieved even when the bit rate is reduced from 2,400 bps to 1,200 bps.