An overview of text-independent speaker recognition: From features to supervectors

By Tomi H. Kinnunen, Haizhou Li. 2010

1,634 citations recorded by the source corpus.

This paper gives an overview of automatic speaker recognition technology, with an emphasis on text-independent recognition. Speaker recognition has been studied actively for several decades. We give an overview of both the classical and the state-of-the-art methods. We start with the fundamentals of automatic speaker recognition, concerning feature extraction and speaker modeling. We elaborate advanced computational techniques to address robustness and session variability. The recent progress from vectors towards supervectors opens up a new area of exploration and represents a technology trend. We also provide an overview of this recent development and discuss the evaluation methodology of speaker recognition systems. We conclude the paper with discussion on future directions.

Subfields: Artificial Intelligence, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Signal Processing

Citations in the corpus

Exploring the Encoding Layer and Loss Function in End-to-End Speaker and Language Recognition System (2018, background): “Moreover, there is no constraint on the lexicon words thus the training utterances and testing segments may have completely different contents [2].”
Generalized End-to-End Loss for Speaker Verification (2017, neutral): “In TD-SV, the transcript of both enrollment and verification utterances is phonetially constrained, while in TI-SV, there are no lexicon constraints on the transcript of the enrollment or verification utterances, exposing a larger variability of phonemes and utterance durations [3, 4].”
A Novel Learnable Dictionary Encoding Layer for End-to-End Language Identification (2018, methodology): “A phonotactic tokenizer can be considered as a dictionary doing hard assignments with top-1 score [1].”
A review of depression and suicide risk assessment using speech analysis (2015, methodology): “Speech parameterisation or feature extraction is the transformation of the raw speech signal into a more abstract representation of the signal with fewer redundancies (Kinnunen and Li, 2009).”
ASVspoof 2021: Towards Spoofed and Deepfake Speech Detection in the Wild (2022, methodology): “An energy-based voice activity detector (VAD) with the recommended threshold settings [38, Section 5.1] is then used to remove segments labeled as non-speech in addition to any speech segments of less than 50 ms duration.”
The ASVspoof 2017 Challenge: Assessing the Limits of Replay Spoofing Attack Detection (2017, background): “Automatic speaker verification (ASV) [1, 2, 3] technology is used in a growing range of applications which require not only robustness to changes in the acoustic environment, but also resilience to intentional circumvention, known as spoofing [4] or, according to ISO/IEC 30107-1:2016(1) standard, pr”
An Overview of Affective Speech Synthesis and Conversion in the Deep Learning Era (2022, background): “Short-time spectral features are correlates of timbre , which captures the “tone” of an utterance and is highly related to the physiological characteristics of the speaker [87]—though it is also affected by affect and phonetic content.”
Deep bottleneck features for i-vector based text-independent speaker verification (2015, methodology): “The minDCF measure is used by the NIST as the primary evaluation figure, since EER uses an arbitrary detection threshold, which is not practically applicable [18].”
Non-parallel voice conversion using i-vector PLDA: towards unifying speaker verification and transformation (2017, background): “There is much further scope to study links between ASV and VC systems.”
Design, analysis and experimental evaluation of block based transformation in MFCC computation for speaker recognition (2012, methodology): “Speaker recognition is a biometric authentication process where the characteristics of human voice are used as the attribute (Kinnunen and Li, 2010; Campbell et al., 2009).”
Deep Speaker Feature Learning for Text-Independent Speaker Verification (2017, background): “Driven by these motivations, many researchers put their effort in ‘feature engineering’ in the past several decades, and new features were proposed occasionally, from perspectives of different knowledge domains [8].”
Spoofing and countermeasures for speaker verification: A survey (2015, background): “The task of an automatic speaker verification (ASV) system is to accept or reject a claimed identity based on a speech sample (Kinnunen and Li, 2010).”
Optimization Algorithms and Applications for Speech and Language Processing (2013, background): “In the early 1990s, Brown et al. [32] of IBM conducted the pioneering work in statistical modeling approaches, establishing a range of IBM models, known prosaically as “Model 1” through “Model 5.””
An Overview of Speaker Identification: Accuracy and Robustness Issues (2011, methodology): “In speaker verification this front-end compensation is augmented by hand-set or channel score normalization strategies [6], [7], [22].”
Speech Emotion Recognition Using Fourier Parameters (2015, background): “Continuous features are important in delivering the emotional cues of speakers [7], [9], [11] and thus have been widely used in speech emotion recognition [1], [7], [22].”
Teacher-Student Training for Text-Independent Speaker Recognition (2018, background): “SR is typically addressed in either text-dependent (TD) [1] or text-independent (TI) [2] manner.”