Dynamic Time Warping

By Alan Bundy, L. Wallen. 1984

How researchers see this paper

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

In Time-Series & Forecasting

Citations in the corpus

MERLOT: Multimodal Neural Script Knowledge Models (2021, methodology): “To estimate timing info for the clean transcript, we align the noisy and cleaned transcripts on a word-by-word level using Dynamic Time Warping [85]; word-word distance is computed using Levenstein distance.”
Mega-TTS: Zero-Shot Text-to-Speech at Scale with Intrinsic Inductive Bias (2023, methodology): “In terms of the pitch distance, we compute the average dynamic time warping (DTW) [40] distances between the pitch contours of ground-truth speech and synthesized speech.”
Benchmark for Skill Learning from Demonstration: Impact of User Experience, Task Complexity, and Start Configuration on Performance (2019, methodology): “Finally, for methods that require time-aligned trajectories, we also warped the demonstrations to be of the same time duration using dynamic time warping (DTW) [23].”
Gesture recognition for human-robot collaboration: A review (2017, background): “Table 6 lists the advantages and disadvantages of the gesture classi ﬁ cation approaches.”
On the information rate of speech communication (2017, methodology): “Additionally, a dynamic time-warping algorithm [21] was applied to all signals in the chorus to ensure that the signals contained the same lexical information at a given time.”
Keystroke Recognition Using WiFi Signals (2015, background): “DTW distance is the Euclidean distance of the optimal warping path between two waveforms calculated under boundary conditions and local path constraints [25].”
EditSpeech: A Text Based Speech Editing System Using Partial Inference and Bidirectional Fusion (2021, methodology): “Then DTW [29] of mel-cepstral coefﬁcient (MCEP) is used to align the unmodiﬁed region of original and candidate mel-spectrogram.”
Any-to-Many Voice Conversion With Location-Relative Sequence-to-Sequence Modeling (2020, methodology): “frame-by-frame using alignment algorithms, such as dynamic time warping (DTW) [1].”
Prosospeech: Enhancing Prosody with Quantized Vector Pre-Training in Text-To-Speech (2022, methodology): “We also use average pitch dynamic time warping (DTW) [23] distance ( D pit ) and duration KL-divergence (KL dur ) as the objective metrics to evaluate the overall prosody: average pitch DTW )) / S W , where W is the Chinese character dictionary, KL ( · ) calculates the KL-divergence, KDE ( · ) is…”
Video Liveness for Citizen Journalism: Attacks and Defenses (2017, methodology): “For each axis, DTW returns the number of diagonal, expansion and contraction moves that convert one vector to the other, and the cost of the resulting transformation.”
Few-Shot Video Classification via Temporal Alignment (2019, methodology): “The video distance measure is thus given by We propose to use a variant of Dynamic Time Warping (DTW) algorithm [24] to solve Eq.”
Effective Use of Variational Embedding Capacity in Expressive End-to-End Speech Synthesis (2019, methodology): “Then we use the dynamic time warping (DTW) algorithm (M¨uller, 2007) (with a warp penalty of 1.0) to ﬁnd an alignment between two spectrograms that produces the minimum MCD cost (including the total warp penalty).”
End-To-End Speech Synthesis Applied to Brazilian Portuguese (2020, methodology): “For the preliminary evaluation, we used the measures L1 loss [Goodfellow et al., 2016], Root Mean Square Error (RMSE) [Tamamori et al., 2017a] and the Dynamic Time Warping (DTW) distance [Müller, 2007].”