The results of some experiments using a discrete cosine transform (DCT) to represent vowel spectra for classification by a neural network are described. The results are compared to a Gaussian classifier trained on the same database. The results show that the DCT classifies vowels using fewer coefficients than the cepstrum. The neural network classifier performs better than the Gaussian classifier, especially with large input feature sets consisting of delta coefficients and formant/pitch features. Best performance using these features was 58.2%. This compares well with other results reported for these data.>

Computer recognition of cursive handwriting is complisated by the lack of letter separation. In an attempt to gain insight on the cursive problem, the simpler problem of recognizing separated hand-written letters is addressed here. A warp-based character matching technique is employed with dictionary lookup, using a 16 000 word vocabulary. Cooperative users achieve high recognition accuracy with this on-line system, which is easily tailored to the individual. Possible extensions to cursive writing are discussed.

Dynamic time warping is a well-established technique for time alignment and comparison of speech and image patterns. This paper decribes the architecture, algorithms, and design of a CMOS integrated processing array used for computing the dynamic time warp algorithm. Emphasis is placed on speech recognition applications because of the real-time constraints imposed by isolated and continuous speech recognition.

Dynamic time warping is an established technique for time alignment and comparison of speech segments in speech recognition. This paper describes a CMOS integrated array processor for computing the dynamic time warp algorithm. It allows many popular variations including LPC and frequency domain representations of speech. High speed is obtained by extensive pipelining, parallel computation, and simultaneous matching of multiple patterns. A realistic application using 40 nine-component LPC vectors per word permits 10,000 word comparisons per second or, equivalently, real time recognition of a 10,000 word vocabulary.

Recently a dynamic elastic model [1] was proposed for automatic image matching. Examples were shown applying the model to dot patterns and gray scale pictures. This paper extends the model to line drawings. Examples are shown on handprint and animation, suggesting the use of dynamic matching for shape recognition and for motion correspondence.