This paper presents the main paradigms for speaker identification, and recent work on missing data methods to increase robustness. The feature extraction, speaker modeling and system classification are discussed. Evaluations of speaker identification performance subject to environmental noise are presented. While performance is impressive in clean speech conditions, there is rapid degradation with mismatched additive noise. Missing data methods can compensate against arbitrary disturbances and remove environmental mismatches. An overview of missing data methods is provided and applications to robust speaker identification summarized. Finally combined approaches involving bottom-up estimation and top-down processing are reviewed, and their significance discussed.

An Overview of Speaker Identification: Accuracy and Robustness Issues

Speaker identity, the sound of a person's voice, plays an important role in human communication. With speech systems becoming more and more ubiquitous, Voice Transformation (VT), a technology that modifies a source speaker's speech utterance to sound as if a target speaker had spoken it, offers a number of useful applications. For example, a novice user can adapt a text-to-speech system to speak with a new voice quickly and inexpensively. 
In this dissertation, we consider new approaches in both the design and the evaluation of VT techniques. We propose a new type of speech corpus that is especially suited to VT research and development by consisting of naturally time-aligned sentences. Consequently, removal of individual prosodic characteristics, such as fundamental pitch and durations, requires only very little processing and results in high-quality speech samples that only differ in their segmental properties, our focus of transformation. These “prosody-normalized” speech samples are used for training VT systems, as well as for evaluating their transformation performance objectively and subjectively. 
Our baseline transformation system (SET) is based on transforming the spectral envelope as represented by the LPC spectrum, using a harmonic sinusoidal model for analysis and synthesis. The transformation function is implemented as a regressive, joint-density Gaussian mixture model, trained on aligned LSF vectors by an expectation maximization algorithm. We improve upon the baseline by adding a residual prediction module, which predicts target LPC residuals from transformed LPC spectral envelopes, using a classifier and residual codebooks. The resulting high resolution transformation system (HRT) is capable of rendering transformed speech with a high degree of spectral detail. 
Because of the severe shortcomings of evaluating VT performance objectively, we propose a subjective evaluation strategy, consisting of several listening tests. In a speaker discrimination test, the HRT system performed significantly better than the SET system. However, discrimination is below that of natural utterances. Similarly, listeners selected the HRT system over other systems in a system comparison test. Finally, listeners rated the speech quality of the HRT system as better than the SET system. However, the quality of natural utterances was considered better than that of transformed speech.

High-resolution voice transformation

A system and method for identifying the phoneme sound types that are contained within an audio speech signal is disclosed. The system includes a microphone (12) and associated conditioning circuitry (14), for receiving an audio speech signal and converting it to a representative electrical signal. The electrical signal is then sampled and converted to a digital audio signal with a digital-to-analog converter (34). The digital audio signal is input to a programmable digital sound processor (18), which digitally processes the sound so as to extract various time domain and frequency domain sound characteristics. These characteristics are input to a programmable host sound processor (20) which compares the sound characteristics to standard sound data. Based on this comparison, the host sound processor (20) identifies the specific phoneme sounds that are contained within the audio speech signal. The programmable host sound processor (20) further includes linguistic processing program methods to convert the phoneme sounds into English words or other natural language words. These words are input to a host processor (22), which then utilizes the words as either data or commands.

User independent, real-time speech recognition system and method

Front-end or feature extractor is the first component in an automatic speaker recognition system. Feature extraction transforms the raw speech signal into a compact but effective representation that is more stable and discriminative than the original signal. Since the front-end is the first component in the chain, the quality of the later components (speaker modeling and pattern matching) is strongly determined by the quality of the front-end. In other words, classification can be at most as accurate as the features. Several feature extraction methods have been proposed, and successfully exploited in the speaker recognition task. However, almost exclusively, the methods are adopted directly from the speech recognition task. This is somewhat ironical, considering the opposite nature of the two tasks. In speech recognition, speaker variability is one of the major error sources, whereas in speaker recognition it is the information that we wish to extract. The mel-frequency cepstral coefficients (MFCC) is the most evident example of a feature set that is extensively used in speaker recognition, but originally developed for speech recognition purposes. When MFCC front-end is used in speaker recognition system, one makes an implicit assumption that the human hearing meachanism is the optimal speaker recognizer. However, this has not been confirmed, and in fact opposite results exist. Although several methods adopted from speech recognition have shown to work well in practise, they are often used as “black boxes” with fixed parameters. It is not understood what kind of information the features capture from the speech signal. Understanding the features at some level requires experience from specific areas such as speech physiology, acoustic phonetics, digital signal processing and statistical pattern recognition. According to the author’s general impression of literature, it seems more and more that currently, at the best we are guessing what is the code in the signal that carries our individuality. This thesis has two main purposes. On the one hand, we attempt to see the feature extraction as a whole, starting from understanding the speech production process, what is known about speaker individuality, and then going

http://cs.joensuu.fi/pages/pums/public_results/2004_PhLic_Kinnunen_Tomi.pdf

Spectral Features for Automatic Text-Independent Speaker Recognition

This paper addresses the issue of closed-set text-independent speaker identification from samples of speech recorded over the telephone. It focuses on the effects of acoustic mismatches between training and testing data, and concentrates on two approaches: (1) extracting features that are robust against channel variations and (2) transforming the speaker models to compensate for channel effects. First, an experimental study shows that optimizing the front end processing of the speech signal can significantly improve speaker recognition performance. A new filterbank design is introduced to improve the robustness of the speech spectrum computation in the front-end unit. Next, a new feature based on spectral slopes is described. Its ability to discriminate between speakers is shown to be superior to that of the traditional cepstrum. This feature can be used alone or combined with the cepstrum. The second part of the paper presents two model transformation methods that further reduce channel effects. These methods make use of a locally collected stereo database to estimate a speaker-independent variance transformation for each speech feature used by the classifier. The transformations constructed on this stereo database can then be applied to speaker models derived from other databases. Combined, the methods developed in this paper resulted in a 38% relative improvement on the closed-set 30-s training 5-s testing condition of the NIST'95 Evaluation task, after cepstral mean removal.

/pdf/robust-text-independent-speaker-identification-over-37tgdjjnhf.pdf

Robust text-independent speaker identification over telephone channels

The aim of the study described in this paper is to provide a thorough and quantitative assessment of the relative speaker discriminating properties of phonemes. A VQ codebook based approach to speaker modeling is used in conjunction with a phonetically hand-labelled database to produce phoneme rankings based on speaker verification scores. In broad groupings the nasals and vowels are found to provide the best speaker recognition performance, followed by the fricatives, affricates and approximants, with the stops providing the worst performance of all. A comparison at the individual phoneme level produces a more detailed ranking and of particular interest is the surprisingly good performance of the unvoiced fricative /s/. The ranking of phonemes is found to be largely unaffected by changes in experimental parameters such as the model size, the feature type and the speaker population. >

J.P. Eatock

Papers

A quantitative assessment of the relative speaker discriminating properties of phonemes