Jared J. Wolf

Methods and experiments for text-independent speaker recognition over telephone channels

We consider methods for text-independent speaker identification that deal with the variability in the data introduced by unknown telephone channels. The methods investigated include probabilistic channel modeling, a channel-invariant model and a modified-Gaussian model. The methods are described and then evaluated with experiments conducted with a twenty speaker database of long distance telephone calls.

Investigation of text-independent speaker indentification over telephone channels

In this paper, we examine several methods for text-independent speaker identification of telephone speech with limited duration data, The issue addressed is the assessment of channel characteristics, especially linear aspects, and methods for improving speaker identification performance when the speaker to be identified is on a different telephone channel than that data used for training. We show experimental evidence illustrating the cross-channel problem and also show that the direct approach, of using simple channel-invariant features, can discard much speaker dependent information. The methods we have found to be most effective rely on the training process to incorporate channel variability.

Investigation of text-independent speaker identification techniques under conditions of variable data

In this paper, we examine several methods for text-independent speaker identification and discuss the problem of achieving robust performance. In a previous paper [1], we have described a radio-channel database that contains highly variable speech data of poor quality. Here, we describe several experiments with the radio-channel database leading to the development of robust features and methods. These experiments show that robustness to one degradation may not be sufficient to improve speaker identification accuracy. A robust feature set and modeling and classification method should mitigate the effects of many of the degradations in the data.

Further investigation of probabilistic methods for text-independent speaker identification

In this paper, we present the preliminary performance of four methods for text-independent speaker identification using speech transmitted over radio channels. In a previous paper [1], we showed that for both laboratory-quality and simulated noisy-channel data in a single-session paradigm, new probabilistic classifiers yielded performance superior to that of a minimum distance classifier. We have recently compiled a speech database consisting of speech transmissions over a radio-channel. The lower quality and higher variability of this database differ markedly from the laboratory-quality databases often used in speech processing research. We present preliminary results with the same four methods of text-independent speaker identification using the radio-channel database with several experimental paradigms including multi-session paradigms. These results show that the probabilistic methods perform significantly better than a minimum-distance classifier for the multi-session paradigm.

Real-Time Speech Coder Implementation on an Array Processor

The feasibility of a speech coding algorithm is most effectively indicated by its successful real-time implementation. The implementation effort presents issues distinct from those related to the development of the algorithm. Problems of real-time software design and debugging, system reliability, and implementation correctness must be addressed. In addition, the power of a general purpose array processing computer may be required to accomplish the real-time aspect of the implementation. The unique, often highly parallel, architecture of available array processors can present the additional issues of multiprocessor intercommunication and control. This paper considers successful approaches to these issues in the context of two real-time speech coder algorithms implemented on an array processor, a linear predictive baseband coder operating at 9.6 kbits/s, and an adaptive predictive coder operating at 16 kbits/s.

Real-time speech coder implementation at 9.6 and 16 kb/s

The practical feasibility of a speech coding algorithm is most effectively indicated by its successful real-time implementation. The implementation effort presents issues distinct from those related to the development of the algorithm. Problems of real-time software design and debugging, system reliability, and implementation correctness must be addressed. In addition, the power of a special purpose array processing computer is generally required to accomplish the real-time aspect of the implementation. The unique, often highly parallel architecture of available array processors can present the additional issues of multi-processor intercommunication and control. This paper considers successful approaches to these issues in the context of two real-time speech coder algorithms implemented on an array processor (a linear predictive baseband coder operating at 9.6 kb/s, and an adaptive predictive coder operating at 16 kb/s).

Speech Signal Processing and Feature Extraction

Speech signal processing and feature extraction is the initial stage of any speech recognition system; it is through this component that the system views the speech signal itself. This chapter introduces general approaches to signal processing and feature extraction and surveys the techniques currently available in these areas.

Organizing a System for Continuous Speech Understanding

In spectrogram reading experiments [Klatt and Stevens, Boston Speech Conference (1972)], the performance obtained by human experts for phonetic segmentation and labeling was approximately 75% partially or completely labeled but correctly specified, 15% mislabeled, and 10% segments missed. When syntactic, semantic, and vocabulary constraints were applied, the success rate in identifying words was 96%. How does one design a computer system to match this performance at word identification? This paper describes the BBN speech understanding system, which is being designed to extract the meaning from continuous speech inputs in a limited semantic environment by using syntactic, semantic, and vocabulary information to overcome vagueness and errors in phonetic segmentation and labeling. The system contains components for acoustic feature extraction, phonetic segmentation and labeling, word retrieval from partial phonetic information, evaluation of word‐match quality, and syntactic and semantic evaluation of parti...

Phonetic and lexical processing in the HWIM speech understanding system

The “front end” of HWIM, the BBN speech understanding system, is that part of the system that governs the formation and evaluation of hypotheses between the levels of the speech signal and the word. It comprises processes for signal processing, acoustic‐phonetic recognition, lexical‐segmental matching, and lexical‐parametric matching. Implicit in the lexical matching processes is the application of phonological rules, both within word pronunciations and across word boundaries. A consistent scoring policy governs the evaluation of hypothesis at the segmental and word levels, and this policy is carried into the control component of the system, where it is applied to multiword hypotheses about the interpretation of the utterance. [This research was supported by the Advanced Research Projects Agency of the Department of Defense and was monitored by ONR under Contract No. N00014‐75‐C‐0053.]

Linguistic processing and control strategy in the HWIM speech understanding system

The principal source of higher‐level linguistic knowledge in HWIM, the BBN speech understanding system, is an augmented transition network parser, which embodies the syntactic, semantic, and part of the factual sources of knowledge of the system. It parses sentences or sentence fragments in either direction, and it can, for a sentence fragment, enumerate the words and syntactic/semantic classes permissible at the ends of the fragment. The control component of the system is a program that calls on the other sources of knowledge of the system in order to formulate, evaluate, and extend hypotheses about the interpretation of the utterance. It is responsible for guiding the system to the most likely interpretation as efficiently as possible. [This research was supported by the Advanced Research Projects Agency of the Department of Defense and was monitored by ONR under Contract No. N00014‐75‐C‐0053.]