State of the art in continuous speech recognition.

Abstract

In the past decade, tremendous advances in the state of the art of automatic speech recognition by machine have taken place. A reduction in the word error rate by more than a factor of 5 and an increase in recognition speeds by several orders of magnitude (brought about by a combination of faster recognition search algorithms and more powerful computers), have combined to make high-accuracy, speakerindependent, continuous speech recognition for large vocabularies possible in real time, on off-the-shelf workstations, without the aid of special hardware. These advances promise to make speech recognition technology readily available to the general public. This paper focuses on the speech recognition advances made through better speech modeling techniques, chiefly through more accurate mathematical modeling of speech sounds.More and more, speech recognition technology is making its way from the laboratory to real-world applications. Recently, a qualitative change in the state of the art has emerged that promises to bring speech recognition capabilities within the reach of anyone with access to a workstation. High-accuracy, real-time, speaker-independent, continuous speech recognition for medium-sized vocabularies (a few thousand words) is now possible in software on off-the-shelf workstations. Users will be able to tailor recognition capabilities to their own applications. Such software-based, real-time solutions usher in a whole new era in the development and utility of speech recognition technology.As is often the case in technology, a paradigm shift occurs when several developments converge to make a new capability possible. In the case of continuous speech recognition, the following advances have converged to make the new technology possible: * higher-accuracy continuous speech recognition, based on better speech modeling techniques; * better recognition search strategies that reduce the time needed for high-accuracy recognition; and * increased power of audio-capable, off-the-shelf workstations. The paradigm shift is taking place in the way we view and use speech recognition. Rather than being mostly a laboratory endeavor, speech recognition is fast becoming a technology that is pervasive and will have a profound influence on the way humans communicate with machines and with each other. This paper focuses on speech modeling advances in continuous speech recognition, with an exposition of hidden Markov models (HMMs), the mathematical backbone behind these advances. While knowledge of properties of the speech signal and of speech perception have always played a role, recent improvements have relied largely on solid mathematical and probabilistic modeling methods, especially the use of HMMs for modeling speech sounds. These methods are capable of modeling time and spectral variability simultaneously, and the model parameters can be estimated automatically from given training speech data. The traditional processes of segmentation and labeling of speech sounds are now merged into a single probabilistic proce...

Towards the Design and Evaluation of ROILA: A Speech Recognition Friendly Artificial Language

“…In order to make our choice we carried out a rationale decision making process by utilizing the Questions, Options and Criteria (QOC) technique [8]. For this purpose we defined the following important criteria for every grammatical property: Learnability; defines whether the grammatical marking in question would be easy to learn or not, Expected recognition accuracy; defines the effect the grammatical marking would have on the anticipated word error rate given that the more constrained a grammar (lower perplexity) is the better it would be for recognition [9], Vocabulary size; describes the effect the grammatical marking would have on increasing or decreasing the vocabulary size, Expressive Ability of the language; defines whether using the grammatical marking in question would actually enable speakers to express more concepts then they would have been unable to do so otherwise, Efficiency; simply relates the grammatical marking to how many words would be required to communicate any solitary meaning, Acknowledgement within Natural and Artificial Languages; states the popularity of the particular grammatical marking amongst each type of languages. Appropriate weights were assigned to the criteria based on importance, for e.g.…”

Emotions, speech and the ASR framework

“…HMMs are used to model the acoustic properties of the recognition tokens, which may be entire words, sub-word units or combinations of words. In the HMM framework, speech is modeled as a two-step probabilistic process (Rabiner and Huang, 1993;Makhoul and Schartz, 1994). In the first step, speech is modeled as a sequence of acoustic states.…”

Classification of time series patterns from complex dynamic systems

“…HMMs are doubly stochastic because both transition and output probabilities come into play. The Viterbi search algorithm is used to find the sequence of states most likely to generate a given output sequence, without searching all possible sequences (Makhoul and Schwartz 1994). Most phonetic HMMs are 3-state left-to-right models as in Figure 5 which represent the beginning, middle and end stages of an utterance.…”