Recurrent input transformations for Hidden Markov models

This paper presents a new architecture which integrates recurrent input transformations (RIT) and continuous density HMMs. The basic HMM structure is extended to accommodate recurrent neural networks which transform the input observations before they enter the Gaussian output distributions associated with the states of the HMM. During training the parameters of both HMM and RIT are simultaneously optimized according to the Maximum Mutual Information (MMI) criterion. Results are presented for the E-set recognition task which demonstrate the ability of recurrent input transformations to exploit longer term correlations in the speech signal and to give improved discrimination.

Effects of pedestal suppression on gain-switched laser sources for 40 Gbit/s OTDM transmission

The performance of 40 Gbit/s optical time-division multiplexed (OTDM) communication systems can be severely limited when the extinction ratio of the optical pulses is low. This is a consequence of the coherent interference noise between individual OTDM channels. When taken alone, the multiple quantum well-distributed feedback laser+dispersion compensating fiber source exhibits a relatively poor extinction ratio which impairs its potential for use in a 40 Gbit/s OTDM system. However, with the addition of an electroabsorption modulator to suppress the pulse pedestals to better than 30 dB extinction, coherent interference noise is reduced, the bit-error-rate performance is greatly improved, and the source shows good potential for 40 Gbit/s OTDM communication.

A comparison of power losses for small (< 1 kW) induction motor drives adopting squarewave and sinusoidal PWM excitation

In this paper the influence of the form of motor excitation on the performance of a small (< 1 kW) induction motor drive is studied. Two forms of excitation, namely sine waves generated by pulse width modulation and simple square wave are explored. Sine wave excitation gives lower motor losses but increases inverter losses. Conversely, square wave excitation increases motor losses but decreases inverter losses. Losses have been measured directly by calorimetric means or, in the case of the inverter, predicted by a Pspice model that has been verified by calorimetric methods. The work shows that overall, the use of square wave excitation leads to a more efficient drive. © 2004 The Institution of Electrical Engineers.