Automatic speaker segmentation and clustering methods have improved considerably over the last few years in the broadcast news domain. However, these generally still produce locally consistent relative labels (such as spkr1, spkr2) rather than true speaker identities (such as Bill Clinton, Ted Koppel). This paper presents a system which attempts to find these true identities from the text transcription of the audio using lexical pattern matching, and shows the effect on performance when using state-of-the-art speaker clustering and speech-to-text transcription systems instead of manual references

Broadcast news (BN) transcription has been a challenging research area for many years. In the last couple of years, the availability of large amounts of roughly transcribed acoustic training data and advanced model training techniques has offered the opportunity to greatly reduce the error rate on this task. This paper describes the design and performance of BN transcription systems which make use of these developments. First, the effects of using lightly supervised training data and advanced acoustic modeling techniques are discussed. The design of a real-time broadcast news recognition system is then detailed using these new models. As system combination has been found to yield large gains in performance, a range of frameworks that allow multiple recognition outputs to be combined are next described. These include the use of multiple types of acoustic models and multiple segmentations. As a contrast a system developed by multiple sites allowing cross-site combination, the "SuperEARS" system, is also described. The various models and recognition configurations are evaluated using several recent BN development and evaluation test sets. These new BN transcription systems can give gains of over 25% relative to the CU-HTK 2003 BN system

Audio diarization is the process of annotating an input audio channel with information that attributes (possibly overlapping) temporal regions of signal energy to their specific sources. These sources can include particular speakers, music, background noise sources, and other signal source/channel characteristics. Diarization can be used for helping speech recognition, facilitating the searching and indexing of audio archives, and increasing the richness of automatic transcriptions, making them more readable. In this paper, we provide an overview of the approaches currently used in a key area of audio diarization, namely speaker diarization, and discuss their relative merits and limitations. Performances using the different techniques are compared within the framework of the speaker diarization task in the DARPA EARS Rich Transcription evaluations. We also look at how the techniques are being introduced into real broadcast news systems and their portability to other domains and tasks such as meetings and speaker verification

Speaker diarisation is the task of automatically segmenting audio data and providing speaker labels for the resulting regions of audio. A cluster-voting scheme is described which takes the output from two speaker diarisation systems and produces a new output which aims to have a lower speaker diarisation error rate (DER) than either input. The scheme works in two stages: the first produces a set of possible outputs which minimise a distance metric based on the DER; the second votes between these alternatives to give the final output. Decisions where the inputs agree are always passed to the output and those where the inputs differ are re-evaluated in the final voting stage. Results are presented on the 6-show RT-03 broadcast news evaluation data; they show that the DER can be reduced by 1.64% and 2.56% absolute using this method when combining the best two Cambridge University and the best two MIT Lincoln Laboratory diarisation systems respectively.

This paper describes the speaker diarisation system developed at Cambridge University in March 2005. This system combines techniques used successfully in our previous speaker diarisation systems with an additional second clustering stage based on state-of-the-art speaker identification methods. Several strategies for using the new system are investigated and the final system gives a diarisation error rate of 6.9 % on the RT-04 Fall diarisation evaluation data when processing all the test data together or 8.6 % when processing the test data shows independently.

Speech recognition systems for conversational telephone speech require the audio data to be automatically divided into regions of speech and non-speech. The quality of this audio segmentation affects the recognition accuracy. This paper describes several approaches to segmentation and compares the resulting recogniser performance. It is shown that using Gaussian mixture models outperforms an energy-detection method and using the output from the speech recogniser itself increases performance further. An upper bound on possible performance was obtained when deriving a segmentation from a forced alignment of the reference words and this outperformed using manually marked word times. Finally the correlation between an appropriately defined segmentation score and WER is shown to be over 0.95 across three data sets, suggesting that segmentations can be evaluated directly without the need for full decoding runs.

A technique for searching audio data to find an exact match for a given piece of cue-audio is described. The method uses a cepstral parameterisation of the audio and a covariance-based distance metric to quickly locate direct repeats. Results on data from ABC news broadcasts show that the method can successfully locate matches several hundred times faster than real-time and requires less than a second of cue-audio. By applying the match recursively to the data, repeated sections of audio, which nearly always correspond to non-news items such as commercials and theme-music, can be identified. Experiments show that the application of the technique can also lead to improved information retrieval using automatically transcribed broadcast data.

No abstract available.

The effects of out-of-vocabulary (OOV) items in spoken document retrieval (SDR) are investigated. Several sets of transcriptions were created for the TREC-8 SDR task using a speech recognition system varying the vocabulary sizes and OOV rates, and the relative retrieval performance measured. The effects of OOV terms on a simple baseline IR system and on more sophisticated retrieval systems are described. The use of a parallel corpus for query and document expansion is found to be especially beneficial, and with this data set, good retrieval performance can be achieved even for fairly high OOV rates.

This paper describes the spoken document retrieval system that we have been developing and assesses its performance using automatic transcriptions of about 50 hours of broadcast news data. The recognition engine is based on the HTK broadcast news transcription system and the retrieval engine is based on the techniques developed at City University. The retrieval performance over a wide range of speech transcription error rates is presented and a number of recognition error metrics that more accurately reflect the impact of transcription errors on retrieval accuracy are defined and computed. The results demonstrate the importance of high accuracy automatic transcription. The final system is currently being evaluated on the 1998 TREC-7 spoken document retrieval task.

Acoustic echo cancellation can be used to remove the annoying talker feedback in hands-free (teleconferencing) systems. The echo canceller identifies the response between the loudspeaker and the microphone, and produces an echo replica which is then subtracted from the signal. Adaptive filtering techniques are employed to determine the echo path response. The speech signal (or the reference signal) is used to train the algorithm. Fast convergence and good tracking capabilities can not be achieved by classical transform domain adaptive filtering algorithms when the reference signal has variable rank autocorrelation matrix. In this thesis, we examine the DCT-LMS algorithm and we emphasize on the role played by the Discrete Cosine Transform. This fixed transformation reduces the eigenvalue spread of the input autocorrelation matrix by partially decorrelating the inputs. The autocorrelation matrix of speech signals is often rank-deficient. During the low rank phases, some of the transform-domain tap coefficients become irrelevant to the adaptation process and stop adapting. When the autocorrelation matrix gains full rank, there will be no longer any “frozen” weights. However, the weights that have been frozen are “far” from the optimal point; they require additional convergence time to track again the changes in the room impulse response. In this dissertation, we present a new method that uses the information contained in the other coefficients to move the frozen weights closer to the optimal point and, consequently, reduce the overall convergence time. By modeling the changes in the impulse response that result from an alteration in the spacing between the microphone and the loudspeaker by a single delay, we were able to develop the “Spectrum Delay Update” method. It consists of replacing, during low-rank phase, each frozen coefficient by a delayed version of the previous full-rank solution. To estimate the corresponding delay, a novel DCT-domain delay estimation algorithm was derived. Simulation results demonstrate the efficiency of SDU for acoustic echo cancellation, the gain in Echo Return Loss is substantial. The experimental performance analysis confirms the expected reduction in the Euclidean Distance between the filter weights and the actual room impulse response DCT. Furthermore, it shows that spectrally updating the filter weights reduces the MSE jump when the autocorrelation matrix gains full rank.

No abstract available.

This paper presents the development of the HTK broadcast news transcription system. Previously we have used data type specific modelling based on adapted Wall Street Journal trained HMMs. However, we are now experimenting with data for which no manual pre-classification or segmentation is available and therefore automatic techniques are required and compatible acoustic modelling strategies adopted. An approach for automatic audio segmentation and classification is described and evaluated as well as extensions to our previous work on segment clustering. A number of recognition experiments are presented that compare datatype specific and non-specific models; differing amounts of training data; the use of gender-dependent modelling and the effects of automatic data-type classification. It is shown that robust segmentation into a small number of audio types is possible and that models trained on a wide variety of data types can yield good performance.

We propose a novel strategy for training neural networks using sequential Monte Carlo algorithms. This global optimisation strategy allows us to learn the probability distribution of the network weights in a sequential framework. It is well suited to applications involving on-line, nonlinear or non-stationary signal processing. We show how the new algorithms can outperform extended Kalman filter (EKF) training.

In this paper speaker clustering schemes are investigated in the context of improving unsupervised adaptation for broadcast news transcription. The various techniques are presented within a framework of top-down split-and-merge clustering. Since these schemes are to be used for MLLRbased adaptation, a natural evaluation metric for clustering is the increase in data likelihood from adaptation. Two types of cluster splitting criteria have been used. The first minimises a covariance-based distance measure and for the second we introduce a two-step E-M type procedure to form clusters which directly maximise the likelihood of the adapted data. It is shown that the direct maximisation technique produces a higher data likelihood and also gives a reduction in word error rate.