Exploiting variable-width features in large vocabulary speech recognition

The use of variable-width features (prosodics, broad structural information etc.) in large vocabulary speech recognition systems is discussed. Although the value of this sort of information has been recognized in the past, previous approaches have not been widely used in speech systems because either they have not been robust enough for realistic, large vocabulary tasks or they have been limited to certain recognizer architectures. A framework for the use of variable-width features is presented which employs the N-Best algorithm with the features being applied in a post-processing phase. The framework is flexible and widely applicable, giving greater scope for exploitation of the features than previous approaches. Large vocabulary speech recognition experiments using TIMIT show that the application of variable-width features has potential benefits.

Chirp limitation on high-speed wavelength conversion using semiconductor optical amplifiers

The cross-gain-saturation effect in SOAs, has been shown to enable robust high-speed wavelength conversion. Under strong electrical and optical pumping, conversion speeds in excess of 20 Gbit/s have been illustrated. However, the effect of chirp on transmission distance at such ultrahigh bit rates has not been studied theoretically in detail. This paper considers the chirp introduced on conversion, employing cross-gain saturation, and studies its dependence on amplifier drive current and signal power. It further shows how an increase in injected cw optical power can reduce chirp while improving conversion speed.

Gain-switched dynamics of tapered waveguide bow-tie lasers

The use of tapered waveguide lasers and amplifiers for enhanced picosecond pulse generation has led to order-of-magnitude peak power and pulse energy improvements. Monolithic pulse generation schemes have so far relied on a double-tapered bow-tie structure. The modeling of tapered lasers has so far been limited to steady-state operation or has lacked experimental comparison. This paper considers both experimentally and theoretically the gain-switched performance of bow-tie lasers of various taper angles. The role of transverse-mode spatial hole burning in tapered waveguide lasers is thereby investigated.

Ultrafast all-optical wavelength conversion using an integrated amplifier/DFB laser

Wavelength conversion in the 1550 nm regime was achieved in an integrated semiconductor optical amplifier (SOA)/DFB laser by modulating the output power of the laser with a light beam of a different wavelength externally injected into the SOA section. A 12 dB output extinction ratio was obtained for an average coupled input power of 75 μW with the laser section driven at 65 mA and the amplifier section at 180 mA. The response time achieved was as low as 13 ps with the laser biased at 175 mA even with low extinction ratios. The laser exhibits a similar recovery time allowing potentially very high bit-rate operation.

High speed wavelength conversion utilising an integrated amplifier/DFB laser

Wavelength conversion in the 1.55-μm regime was achieved for the first time in an integrated SOA/DFB laser by modulating the output power of the laser with a light beam of a different wavelength externally injected into the SOA section. In terms of speed, response times as low as 13ps were observed, though at the expense of reduced extinction ratio. Generally, these results indicate that operation in the 10s of GB/s should be possible.

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.

Wavelength conversion based on an integrated semiconductor optical amplifier/DFB laser at 10 Gb/s with low input power

A study of the relative performance of an integrated semiconductor optical amplifier (SOA)/distributed feedback laser wavelength converter that can operate with negative penalties at 10 Gb/s rates is conducted. It is found that reduction of more than 25 times in required input powers are achieved when compared with laser or SOA converters.

Demonstration of time division demultiplexing from 40Gb/s to 10Gb/s via polarisation switching in a semiconductor optical amplifier

An all-optical polarization rotation technique was demonstrated for demultiplexing a 40 Gb/s return-to-zero optical time division de/multiplexing (OTDM) signal. A sensitivity penalty of 3.5 dB was achieved for the total multiplexing/demultiplexing process from 10Gb/s to 40 Gb/s and back again.

Spatial emission control of vertical cavity surface emitting lasers to provide bandwidth gain in multimode fibre links using a simple alignment technique

Etched VCSEL sources are reported which avoid bandwidth collapse in multimode fibre using a simple coupling technique to control the launch. These devices have allowed better than over-filled launch bandwidth for alignment tolerances of ±7 microns.

All-optical 3R regeneration and wavelength conversion in an integrated SOA/DFB laser: Experiment and simulation

The simultaneous all optical 3R regeneration and format conversion in a simple, single integrated device was examined. The integrated device consisted of a semiconductor optical fiber (SOA) monolithically integrated with a distributed feedback (DFB) laser. Gain saturation was employed for the transmission of a data signal regenerated all-optically in the laser/amplifier device. The regeneration of the electrically filtered eye diagrams was observed by noise removal and extinction ratio-improvement by the device.

10-Gbits/s all-optical 3R regeneration and format conversion using a gain-switched DFB laser

A strain-compensated multiple quantum well device is used as a DFB laser, this has been optimized for low jitter gain switched operation at 10 GHz. The signal is transmitted down 80 km of standard fiber then amplified, filtered and polarization controlled before being injected into a DFB laser. The purpose of this regeneration process is to gain switch the DFB with the extracted clock signal in order to retime the converted signal. This process also simultaneously converts the input NRZ format to an output RZ data to format and results in a signal whose optical power and extinction ratio are considerably improved by the regeneration process.

Using a semiconductor optical amplifier integrated with a pump laser for mid-span spectral inversion and wavelength conversion

A semiconductor optical amplifier monolithically integrated with a distributed feedback pump laser is used for non-degenerate four wave mixing applications. Experimental results are presented which illustrate the use of this compact device for both wavelength conversion and dispersion compensation applications at high data rates.

Adaptive control of combustion instabilities for unknown sign of the high frequency gain

A type of adaptive, closed-loop controllers known as self-tuning regulators present a robust method of eliminating thermoacoustic oscillations in modern gas turbines. These controllers are able to adapt to changes in operating conditions, and require very little pre-characterisation of the system. One piece of information that is required, however, is the sign of the system's high frequency gain (or its 'instantaneous gain'). This poses a problem: combustion systems are infinite-dimensional, and so this information is never known a priori. A possible solution is to use a Nussbaum gain, which guarantees closed-loop stability without knowledge of the sign of the high frequency gain. Despite the theory for such a controller having been developed in the 1980s, it has never, to the authors' knowledge, been demonstrated experimentally. In this paper, a Nussbaum gain is used to stabilise thermoacoustic instability in a Rijke tube. The sign of the high frequency gain of the system is not required, and the controller is robust to large changes in operating conditions - demonstrated by varying the length of the Rijke tube with time. Copyright © 2008 by Simon J. Illingworth & Aimee S. Morgans.

Gaussian mixture modeling with Gaussian process latent variable models

Density modeling is notoriously difficult for high dimensional data. One approach to the problem is to search for a lower dimensional manifold which captures the main characteristics of the data. Recently, the Gaussian Process Latent Variable Model (GPLVM) has successfully been used to find low dimensional manifolds in a variety of complex data. The GPLVM consists of a set of points in a low dimensional latent space, and a stochastic map to the observed space. We show how it can be interpreted as a density model in the observed space. However, the GPLVM is not trained as a density model and therefore yields bad density estimates. We propose a new training strategy and obtain improved generalisation performance and better density estimates in comparative evaluations on several benchmark data sets. © 2010 Springer-Verlag.