James L. Melsa

The relationship between an adaptive quantizer and a variance estimator (Corresp.)

In this correspondence, it is shown that an adaptive quantizer with a one word memory can be viewed as one that estimates the variance of its input and normalizes the input by the square root of the estimate. It is shown that, even though the estimate is an exponential average, the effect of transmission errors does not die out. Finally, a method of combating the effect of such errors is described.

Specific optimal estimation

The goal of the specific optimal estimation problem is to achieve near-optimal (minimum variance) estimation using a structure that is easier to implement than the optimal solution. One chooses a reasonable configuration for the filter in which certain parameters are unspecified and then selects the parameters so that its performance is optimized. The problem is formulated as a two-point boundary-value problem resulting from consideration of the covariance of error of the estimate and application of the matrix-minimum principle. The examples presented indicate that near-optimal results can be obtained using a filter designed in this way.

Unified development of algorithms used for linear predictive coding of speech signals

The requirement for low data rate voice transmission has resulted in a large number of algorithms being proposed for speech digitization at data rates of 2·4–4 kilobits/sec. Many of the proposed algorithms are quite complicated and have their origin in disciplines generally considered to be outside of the realm of the speech researcher or communication system designer. Additionally, the algorithms have been developed and presented in highly varying notation using various theoretical approaches. The result is a confusing array of equations, algorithms, and numerical analysis procedure. It is the goal of this paper to alleviate this problem by providing a unified tutorial development of the various algorithms used and proposed for speech data compression. Classical least squares estimation theory is used as the focal point of the discussion since it forms the basis for several of the more familiar speech digitization algorithms. The remainder of the algorithms, whether they have their basis in stochastic estimation theory or statistical regression theory, are related back to the more familiar least squares approach. The speech digitization techniques discussed are the covariance method, the autocorrelation method, the PARCOR method, a priori analysis, the sequential least squares method, the Kalman filter approach, the stochastic approximation method, and the general linear regression model. An effort has been made to provide sufficient theoretical background to establish the algorithm relationships without stressing mathematical rigor.

Energy savings for a solar heated and cooled building through adaptive optimal control

This paper describes a study of the applicability of adaptive and optimal control techniques to the control of heating, ventilating, and air conditioning (HVAC) systems of solar heated and cooled buildings. The suitability of optimal and adaptive concepts is discussed and the selected approach is explained. An integral quadratic cost functional to define optimal performance and an identification process to produce a linearized building model are combined to yield an adaptive linear regulator solution. The building model is described and heating system simulations of three versions of the adaptive optimal controller are reported along with a simulation of a conventional controller for comparison. A nonlinear, open-loop, optimal control simulation is also reported and used to indicate an upper bound on achievable performance. Cooling system simulation results are also reported for an adaptive optimal controller and a conventional controller. Substantial savings in auxiliary energy requirements are demonstrated by the adaptive optimal controllers.

Mediumband speech encoding using time domain harmonic scaling and adaptive residual coding

This paper describes the study of a new approach to speech digitization at mediumband bit rates of 9.6 to 16 Kb/s. The technique is based on a combination of Time Domain Harmonic Scaling and an Adaptive Residual Coder. Computer simulation studies have shown that this technique is able to produce excellent quality speech at the bit rates in question.

Practical considerations for variable length source coding

Variable length source coding schemes offer substantial improvements in data rate for a wide variety of data compression techniques. Unfortunately, practical considerations have limited their use. This paper describes several techniques which allow these powerful codes to be effectively applied. In particular, the problem of loss of code word synchronization due to channel errors is addressed. It is shown how self-synchronizing codes limit error propagation. Also, variable delay and the associated difficulties of buffer synchronization and buffer overflow are considered. The variable delay at the transmitter can be coupled with a variable delay at the receiver to give a fixed system delay. Buffer control then becomes the allocation of this delay between transmitter and receiver. Finally, a switching property of variable length codes to better match the bimodal operation of the source is presented.

A pitch compensating quantizer

A new adaptive quantizer for speech digitization has been derived. It is similar to known adaptive quantizers in that it adjusts its dynamic range to match that of the speech waveform. In addition, it further adjusts its range to compensate for the increased signal strength that follows a pitch pulse. The new quantizer bases its adaptation on its own output and no side information is required. When combined with a variable length source coding scheme, the new quantizer offers a significant improvement in signal-to-noise ratio and in subjective speech quality. The technique is applicable to a broad range of digitization methods including adaptive delta modulation and various forms of ADPCM.

Custom VLSI design of a single chip multi-channel ADPCM processor

This paper details the development of a single chip VLSI processor which provides fully compatible CCITT 32kbps ADPCM transcoding. The chip is designed for multi-channel systems and is capable of processing eight independent channels in a 125 microsecond frame. To achieve this processing power, a specialized microcoded architecture for the ADPCM algorithm is implemented using a full custom two micron VLSI design methodology.

CVSD to LPC conversion using noise tolerant analysis

Due to the increasing sophisitication and decreasing cost of digital hardware, an all-digital speech communication network is being considered for implementation. Subject to bandwidth limitations, present plans call for both 16 Kilobit Continuously Variable Slope Delta Modulation (CVSD) as well as 2.4 Kilobit Linear Predictive Coding (LPC) terminals. Hence the conversion of CVSD to LPC and vice versa arises naturally in this environment. This paper will focus on the CVSD to LPC conversion problem. After a brief discussion of the component system environments, a structure for the format conversion system will be proposed. Since previous work has shown that pitch and voicing decisions can be made on the CVSD observations, the presentation will focus on the identification of LPC predictor coefficients from the noisily observed CVSD data. The remainder of the paper centers on this and the coefficient correction problem, and a new class of linear estimators is shown to yield conditionally unbiased estimates of the LPC coefficients in both noisy and noiseless environments.

A new configuration for speech digitization at 9600 bits per second

A speech coding algorithm for digital transmission of speech at a rate of 9600 bits per second which can be implemented on a speech processing system is described. The algorithm combines the following: a pitch extraction loop, a pitch compensating adaptive quantizer, a sequentially adaptive linear predictor, an adaptive source coding, and a multipath tree searching to generate very high quality speech output. Although each of these elements has been previously applied to speech coding, the combination of all five of these elements has not been discussed before. Preliminary simulation studies of the algorithm have yielded a signal-to-noise ratio which is indicative of very high quality speech.