Pith. sign in

REVIEW 4 major objections 5 minor 31 references

Streaming Keyword Spotting Boosted by Cross-layer Discrimination Consistency

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Keyword spotting improves when the decoder checks that two network layers agree on the wake word.

desk verdict Solid streaming CTC decoder, but the CDC boost is largely a look-ahead effect that the paper never isolates. read the letter →

arxiv 2412.12635 v2 pith:GUUK4HF3 submitted 2024-12-17 eess.AS cs.SD

Yu Xi , Haoyu Li , Xiaoyu Gu , Hao Li , Yidi Jiang , Kai Yu This is my paper · ORCID
classification eess.AScs.SD
keywords keywordspottingwakeworddetectionCTCdecodingstreamingintermediateregularizationcross-layerconsistencyfalsealarmreductionHeySnips
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that a keyword-spotting system trained with Connectionist Temporal Classification can decode a wake word in a live audio stream more accurately if the decoder restricts its search to the keyword's phoneme sequence and then checks that the network's intermediate and final layers agree on what they are hearing. The proposed streaming algorithm detects a keyword start at any frame, unlike ASR-style decoding that searches the whole vocabulary from the beginning of an utterance, and it avoids the engineering overhead of graph-based decoders. On the Hey Snips wake word "Hey Snips," the authors report that adding the cross-layer consistency check raises average recall by 6.8 percentage points and cuts the miss rate by 46.3% relative to a graph-based CTC baseline, at a false alarm rate of 0.05 per hour. If this holds, it offers a simple path to better on-device wake word detection, especially in noisy conditions.

What carries the argument

The mechanism is a two-stage streaming decoder. Stage one (Algorithm 1) runs a frame-synchronous search over the CTC posterior matrix, but the search alphabet is restricted to the keyword phoneme sequence with blanks inserted ($\tilde y = [\phi, y_1, \phi, \ldots, \phi, y_U, \phi]$); at every frame a new path competitor is initialized, paths longer than a timeout are discarded, and each surviving score is normalized by path length. Stage two defines a cross-layer discrimination consistency (CDC) score: for each frame $t$, take a sliding window of the score curves $s^{\mathrm{init}}$ and $s^{\mathrm{inter}}$ produced by the final and intermediate CTC branches, compute their cosine similarity, and set the refined score to $(s^{\mathrm{init}} + s^{\mathrm{cdc}})/2$. The CDC score carries the information that sustained agreement between the shallow and deep branches indicates a true keyword, while a brief, divergent spike indicates a false alarm.

What would settle it

Evaluate the CDC-boosted decoder on a confusable-phrase test set (utterances that sound similar to "Hey Snips" but are not the wake word) under noise types not seen in training, such as babble or music; if average recall at 0.05 false alarms per hour falls back to the non-CDC streaming level, the cross-layer consistency signal is not carrying the discrimination.

Watch

Extended reading notes

Core claim

The central claim is that a frame-synchronous decoding algorithm, which tracks only partial keyword hypotheses in the CTC posterior stream and can start a new path at any time step, yields higher keyword recall than both ASR-style decoding (greedy search and prefix beam search) and a WFST graph-based decoder using the same acoustic model. The further claim is that false alarms can be suppressed by exploiting a behavioral difference between true and spurious activations: near a genuine wake-up, the frame-level scores from the intermediate CTC branch and the final CTC branch stay stable and close to each other, whereas for false alarms the scores spike briefly and diverge. The paper defines a cross-layer discrimination consistency (CDC) score as the cosine similarity between the two score curves over a sliding window, averages it with the initial CTC score, and shows that this refined score improves recall at low false alarm rates, with the largest gains at low signal-to-noise ratios. The performance numbers are reported on clean and WHAM!-noised Hey Snips data: the CDC-boosted streaming decoder reaches 92.1% average recall at 0.05 false alarms per hour, up from 85.3% for the graph-based baseline.

Load-bearing premise

The load-bearing assumption is that a high cosine similarity between the intermediate and final layer's wake-word score curves over a short sliding window separates true wake words from false alarms reliably across speakers and noise conditions; the paper supports this with one illustrative example and a window-size sweep rather than a systematic analysis.

Editorial extensions

If this is right

  • The same CTC acoustic model, decoded with the proposed streaming algorithm instead of ASR greedy or prefix beam search, gains 10.4% and 6.6% absolute accuracy, respectively, at a false alarm rate of zero.
  • Adding the CDC refinement cuts the average miss rate by 46.3% relative to the graph-based baseline at 0.05 false alarms per hour, with the largest absolute gains at -5 dB and 0 dB SNR.
  • A look-ahead of about 900 ms (30 future frames) is sufficient to reach near the performance of an offline decoder that sees the whole utterance.
  • The best CDC behavior comes from placing the intermediate CTC branch at roughly half the encoder depth (layer 3 of 6) and setting the intermediate loss weight to $w = 0.3$.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The cross-layer agreement check is not tied to CTC or to monophone units; the same consistency statistic between shallow and deep representations could gate detections in other streaming models such as RNN-T or attention-based wake word systems, though the paper does not test this.
  • Because the decoder confines each search to one keyword, running several keyword-specific searches in parallel could support multi-keyword or open-vocabulary spotting without a decoding graph, an extension the paper does not evaluate.
  • The separation between true and false activations is only as good as the variety of false alarms seen during development; a natural stress test would be to measure whether the CDC score still separates classes on confusable phrases or noise types not present in the WHAM! corpus.
  • The reported 6.8-point recall gain is measured on a single wake word ("Hey Snips") and a single noise corpus; a principled follow-up would be to repeat the comparison on a second keyword or another language to see whether the gain persists.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper proposes a decoding method for CTC-based streaming keyword spotting (KWS). The first component (Section II-B, Algorithm 1) is a frame-synchronous Viterbi-style search over the keyword's phoneme sequence with blanks inserted; new path competitors are initialized at every frame, a bonus and a timeout control the search, and the output is a per-frame detection score. This decoder is compared against greedy, prefix-beam, and WFST-graph CTC decoders using the same acoustic model. The second component (Section II-C) is a refinement stage that computes the cosine similarity between the score curves of an intermediate CTC branch and the main CTC branch over a sliding window (Equation 5) and averages it with the initial score (Equation 6). On the Hey-Snips dataset with WHAM! noise at SNRs from -5 dB to clean, the streaming decoder improves average accuracy over all three baselines (Table I), and with ICTC plus the CDC refinement, configured with a 30-frame future window, the system reaches 92.1% average recall versus 85.3% for the graph baseline at FAR = 0.05/hr, corresponding to a 6.8% absolute recall gain and a 46.3% relative miss-rate reduction (Table II). The paper also sweeps the history/future window sizes (Table III) and the ICTC weight and layer (Table IV).

Significance. If the claims hold, the streaming decoder is a simple and effective alternative to WFST-based graph decoding for CTC-based KWS, and the multi-stage refinement idea is a useful engineering contribution for noisy wake-word detection. The paper has concrete strengths: Algorithm 1 is specified precisely enough to re-implement; the experiments use public corpora (LibriSpeech, Hey-Snips, WHAM!); Table III is a systematic sweep of the latency/accuracy trade-off with clearly labeled streaming and offline upper-bound configurations; and the paper explicitly acknowledges that the future-window size trades latency against accuracy. The headline claims are falsifiable and are stated with their operating point (FAR = 0.05/hr). The weaknesses are concentrated in three areas: the CDC stage's incremental gain is confounded with look-ahead, the results are point estimates without variance or significance assessment, and the evaluation covers a single keyword. Note also that the CDC feature is computed from the same model's own intermediate and final scores, so it cannot serve as independent verification of detections; this is acceptable for a rescoring heuristic but should be described as such.

major comments (4)
  1. [§II-C, Eq. (5); Tables II–III] The incremental benefit claimed for the CDC refinement is confounded with look-ahead. Configuration E in Table II uses L_Fut = 30 frames (900 ms of future context in the sliding-window cosine similarity of Equation (5)), while configuration D uses no future context. Table III shows that the CDC system's average recall grows monotonically with L_Fut (90.7 at L_Fut = 0, 91.0 at L_Fut = 10, 92.1 at L_Fut = 30) and that at L_Fut = 0 it is below the no-CDC streaming system D (91.1 in Table II). Because no equal-latency no-CDC baseline is reported (for example, delaying or window-smoothing s_init alone over the same 900 ms window), the 1.0-point E-versus-D gain in Table II cannot be attributed to cross-layer discrimination consistency; it may be entirely an effect of the additional future context. The statement in Section IV-B that 'the comparison between models (E vs. D) further demonstrates the effectiveness of our proposed CDC strategy' is therefore not supported as written. The missing ablation should be added, and the L_Fut = 0 below-baseline result in Table III should be analyzed explicitly.
  2. [§III-C, §IV-B] All quantitative claims rest on single-run point estimates without error bars, confidence intervals, or significance tests. At FAR = 0.05/hr over a negative set of roughly 97 hours, threshold selection permits only about five false alarms, so both the operating-point fixing and the resulting recall estimates are high-variance; a difference such as D = 91.1 versus E = 92.1 corresponds to about 26 utterances out of 2,599 positives and may lie within sampling noise. The authors should report bootstrap confidence intervals or multiple runs and state explicitly how the FAR threshold is selected (on which subset and with which decision rule). Without this, the abstract's headline numbers (6.8% absolute recall gain, 46.3% relative miss-rate reduction) exceed what the evidence supports.
  3. [§III-B, §IV-C, §IV-D] The hyperparameter analysis does not document a held-out validation protocol. Tables III and IV sweep L_His/L_Fut, the ICTC weight w, and the ICTC layer and report test-set numbers; if these configurations were selected on the test set, the chosen operating point (L_Fut = 30, w = 0.3, layer 3) is optimistically biased. The dev set (2,484 keyword utterances) is available and should be used for selection, or the paper should state that the reported numbers are test-set-tuned. In addition, the decision procedure that converts the score sequence Scores[1:T] into a detection (thresholding, peak-picking, and the role of S_bonus and T_out) is never described, which makes the recall-at-fixed-FAR evaluation irreproducible.
  4. [§III-A, §II-C, Fig. 2] All experiments use a single keyword ('Hey Snips') and a single noise source (WHAM!). The general claims about the streaming decoder and, especially, the CDC observation that intermediate-layer and final-layer scores behave differently for positives and negatives rest on this one keyword, and Figure 2 illustrates the effect with one positive and one negative example. Since the score dynamics of Algorithm 1 depend on the keyword's phoneme sequence and confusability structure, and since the discriminative power of the cosine-similarity feature is an empirical property of the trained model, at least one additional keyword should be evaluated before the mechanism claims in Section I contribution (2) and the conclusion are generalized.
minor comments (5)
  1. [§II-B, Algorithm 1] The pseudo-code has several typos: line 4 reads 'for t = 2or T' (presumably '2 to T'), lines 11 and 14 use 'f rd' (presumably 'fwd'), and the condition 'if ũ = 1 or ũ = 2' at line 5 refers to ũ before the loop variable is introduced; the intended initialization δ(t,1) = δ(t,2) = 1 should be written as part of the t-loop preamble.
  2. [§II-B, Algorithm 1, lines 14–16] The non-blank recurrence allows a transition from ũ−2 without imposing the CTC same-label constraint (ỹ_{ũ−2} ≠ ỹ_ũ). This is harmless for 'Hey Snips', whose monophone sequence has no adjacent repeats, but the condition should be stated for the algorithm to be correct for arbitrary keywords with repeated phonemes.
  3. [§II-C, Eq. (6)] The refinement averages s_init with s_cdc, but the two quantities are not on the same scale: s_init is a probability-like geometric-mean score scaled by S_bonus = e^3 and can exceed 1, while s_cdc is a cosine similarity in [0,1]. A sentence on score normalization, or a plot of the two score distributions, would make the combination in Equation (6) less arbitrary.
  4. [§I, §II-B] Algorithm 1 is a frame-synchronous token-passing search over a keyword-specific phone sequence, a family with earlier representatives in the keyword-search literature; the paper should position the algorithm against frame-synchronous phoneme-alignment search methods (e.g., [12]) and prior CTC-based KWS decoding (e.g., [6]–[8]) rather than only against whole-utterance ASR decoding and WFST graphs, since contribution (1) in Section I depends on that distinction.
  5. [§II-C, Fig. 1, Fig. 2] Notation and presentation: 'LHis.' and 'LFut.' use periods that are easy to misread as sentence punctuation (suggest L_hist/L_fut); Figure 1's caption spells 'consistency' as 'consitency'; Figure 2's abscissa is labeled 'Frames' although each frame represents 30 ms after the frame-skipping of Section III-B; and the claim in Section II-C that score calculation is 'fully streaming' should be qualified, because the refined score at frame t is not final until t + L_Fut.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity; the CDC-vs-look-ahead attribution is a confound, not a circular derivation.

full rationale

The central streaming-decoding claim is self-contained: Algorithm 1 is a Viterbi-style keyword-constrained search over the CTC posterior matrix, and Table I compares it against independent ASR and graph-based baselines. No equation feeds a fitted value back into the reported metric, and no prediction is defined as the quantity it claims to predict. The CDC refinement in Eqs. 5-6 is a fixed feature combination, cosine similarity of the main and intermediate CTC scores; it is not a parameter fit to test labels, and using the same model's outputs to form a score is ordinary inference, not circularity. The only self-citation is to the authors' prior TDT-KWS work [14], used for inspiration and for the Snips negative-set reorganization; it is not load-bearing. A genuine scientific weakness, outside circularity, is that the E-vs-D comparison in Table II attributes a 1.0 recall gain to CDC while D is zero-latency and E uses L_Fut=30 (900 ms). Table III shows the L_Fut=0 CDC system (90.7) falls below the no-CDC streaming row (91.1), so the gain is confounded with look-ahead rather than cleanly attributable to cross-layer consistency. That is a control/attribution risk, not a circular step.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central claims depend on a standard CTC model, an ICTC branch, a hand-designed decoder, and a heuristic CDC score. There are no released artifacts. The main free parameters are decoding hyperparameters and ICTC/CDC configuration, all selected with limited ablations, some on the test set. No new physical or architectural entities are introduced beyond the CDC score function.

free parameters (6)
  • SBonus = e^3
    Bonus score added at keyword end in Algorithm 1, Section III-C; fixed by hand, not learned.
  • Tout = 3 seconds
    Timeout to discard paths longer than the keyword, Section III-C; manually set.
  • ICTC loss weight w = 0.3
    Weight in Eq. 2; selected by sweep in Table IV, not by theory.
  • ICTC application layer = layer 3 of 6 (L/2)
    Position of intermediate CTC branch; selected by sweep in Table IV.
  • CDC history/future window = LHis=0, LFut=30 frames (900 ms latency)
    Window for cosine similarity; best configuration chosen from Table III.
  • Refinement averaging coefficient = 1/2 in Eq. 6
    Arbitrary combination of initial and CDC scores; no ablation over this coefficient is reported.
assumptions (6)
  • domain assumption CTC model outputs per-frame phoneme posteriors and the keyword can be encoded as a blank-inserted phoneme sequence.
    Algorithm 1 relies on this representation; Section II-B.
  • domain assumption CMU Pronouncing Dictionary phoneme sequence for "Hey Snips" is correct and sufficient.
    Training and decoding use cmudict-0.7b; Section III-B.
  • ad hoc to paper Cosine similarity of sliding-window score curves separates positives from negatives.
    Proposed in Section II-C based on Figure 2; no independent derivation.
  • domain assumption Intermediate CTC regularization makes the middle-layer branch informative for keyword discrimination.
    Eq. 2 and Section IV-D; standard ICTC assumption.
  • domain assumption The synthesized noisy test set and 97-hour negative pool yield a fair FAR=0.05/hour evaluation.
    Section III-C; no external benchmark for noisy Snips is used.
  • standard math CTC loss in Eq. 1 is computed over all alignments, and Algorithm 1's Viterbi-style max approximation is valid.
    Follows from Graves et al. [5]; no formal proof is given in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Streaming Keyword Spotting Boosted by Cross-layer Discrimination Consistency." pith.science (2026). https://pith.science/paper/GUUK4HF3

@misc{pith2026241212635,
  author       = {Pith},
  title        = {Pith review of: Streaming Keyword Spotting Boosted by Cross-layer Discrimination Consistency},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GUUK4HF3}},
  note         = {Machine review of arXiv:2412.12635}
}
read the original abstract

Connectionist Temporal Classification (CTC), a non-autoregressive training criterion, is widely used in online keyword spotting (KWS). However, existing CTC-based KWS decoding strategies either rely on Automatic Speech Recognition (ASR), which performs suboptimally due to its broad search over the acoustic space without keyword-specific optimization, or on KWS-specific decoding graphs, which are complex to implement and maintain. In this work, we propose a streaming decoding algorithm enhanced by Cross-layer Discrimination Consistency (CDC), tailored for CTC-based KWS. Specifically, we introduce a streamlined yet effective decoding algorithm capable of detecting the start of the keyword at any arbitrary position. Furthermore, we leverage discrimination consistency information across layers to better differentiate between positive and false alarm samples. Our experiments on both clean and noisy Hey Snips datasets show that the proposed streaming decoding strategy outperforms ASR-based and graph-based KWS baselines. The CDC-boosted decoding further improves performance, yielding an average absolute recall improvement of 6.8% and a 46.3% relative reduction in the miss rate compared to the graph-based KWS baseline, with a very low false alarm rate of 0.05 per hour.

Figures

Figures reproduced from arXiv: 2412.12635 by the authors.

Figure 1
Figure 1. An example of ”Hey Snips” illustrates the training and decoding [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The patterns of frame-level decoding scores are shown for positive [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 13 canonical work pages

  1. [1]

    Small-footprint keyword spotting using deep neural networks,

    G. Chen, C. Parada, and G. Heigold, “Small-footprint keyword spotting using deep neural networks,” in Proc. IEEE ICASSP , 2014, pp. 4087– 4091

    2014

  2. [2]

    End-to-end streaming keyword spotting,

    R. Alvarez and H.-J. Park, “End-to-end streaming keyword spotting,” in Proc. IEEE ICASSP , 2019, pp. 6336–6340

    2019

  3. [3]

    Wake word detection with streaming transformers,

    Y . Wang, H. Lv, D. Povey, L. Xie, and S. Khudanpur, “Wake word detection with streaming transformers,” in Proc. IEEE ICASSP , 2021, pp. 5864–5868

    2021

  4. [4]

    Deep spoken keyword spotting: An overview,

    I. L ´opez-Espejo, Z.-H. Tan, J. H. L. Hansen, and J. Jensen, “Deep spoken keyword spotting: An overview,” IEEE Access, pp. 4169–4199, 2022

    2022

  5. [5]

    Connec- tionist temporal classification: labelling unsegmented sequence data with recurrent neural networks,

    A. Graves, S. Fern ´andez, F. J. Gomez, and J. Schmidhuber, “Connec- tionist temporal classification: labelling unsegmented sequence data with recurrent neural networks,” in Proc. ICML, 2006, pp. 369–376

    2006

  6. [6]

    Unrestricted vocabulary keyword spotting using lstm-ctc

    Y . Zhuang, X. Chang, Y . Qian, and K. Yu, “Unrestricted vocabulary keyword spotting using lstm-ctc.” in Proc. ISCA Interspeech , 2016, pp. 938–942

    work page 2016

  7. [7]

    CRNN-CTC based mandarin keywords spotting,

    H. Yan, Q. He, and W. Xie, “CRNN-CTC based mandarin keywords spotting,” in Proc. IEEE ICASSP , 2020, pp. 7489–7493

    work page 2020

  8. [8]

    Multi-task network for noise- robust keyword spotting and speaker verification using CTC-based soft V AD and global query attention,

    M. Jung, Y . Jung, J. Goo, and H. Kim, “Multi-task network for noise- robust keyword spotting and speaker verification using CTC-based soft V AD and global query attention,” in Proc. ISCA Interspeech, 2020, pp. 931–935

    work page 2020

Show all 31 references
  1. [9]

    Improving rnn transducer modeling for small-footprint keyword spotting,

    Y . Tian, H. Yao, M. Cai, Y . Liu, and Z. Ma, “Improving rnn transducer modeling for small-footprint keyword spotting,” in Proc. IEEE ICASSP, 2021, pp. 5624–5628

    2021

  2. [10]

    U2-KWS: Unified two-pass open-vocabulary keyword spotting with keyword bias,

    A. Zhang, P. Zhou, K. Huang, Y . Zou, M. Liu, and L. Xie, “U2-KWS: Unified two-pass open-vocabulary keyword spotting with keyword bias,” in Proc. IEEE ASRU, 2023, pp. 1–8

    2023

  3. [11]

    CTC-aligned audio-text embedding for streaming open-vocabulary keyword spotting,

    S. Jin, Y . Jung, S. Lee, J. Roh, C. Han, and H. Cho, “CTC-aligned audio-text embedding for streaming open-vocabulary keyword spotting,” CoRR, 2024

    2024

  4. [12]

    Keyword search using attention-based end-to-end asr and frame-synchronous phoneme alignments,

    R. Yang, G. Cheng, H. Miao, T. Li, P. Zhang, and Y . Yan, “Keyword search using attention-based end-to-end asr and frame-synchronous phoneme alignments,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 29, pp. 3202–3215, 2021

    2021

  5. [13]

    WeKWS: A production first small-footprint end-to-end keyword spot- ting toolkit,

    J. Wang, M. Xu, J. Hou, B. Zhang, X. Zhang, L. Xie, and F. Pan, “WeKWS: A production first small-footprint end-to-end keyword spot- ting toolkit,” in Proc. IEEE ICASSP , 2023, pp. 1–5

    2023

  6. [14]

    TDT-KWS: Fast and accurate keyword spotting using token-and-duration transducer,

    Y . Xi, H. Li, B. Yang, H. Li, H. Xu, and K. Yu, “TDT-KWS: Fast and accurate keyword spotting using token-and-duration transducer,” inProc. IEEE ICASSP, 2024, pp. 11 351–11 355

    2024

  7. [15]

    First-pass large vocabulary continuous speech recognition using bi-directional recurrent dnns,

    A. L. Maas, A. Y . Hannun, D. Jurafsky, and A. Y . Ng, “First-pass large vocabulary continuous speech recognition using bi-directional recurrent dnns,” CoRR, vol. abs/1408.2873, 2014. [Online]. Available: http://arxiv.org/abs/1408.2873

    2014 arXiv

  8. [16]

    Compressed time delay neural network for small-footprint keyword spotting

    M. Sun, D. Snyder, Y . Gao, V . K. Nagaraja, M. Rodehorst, S. Pan- chapagesan, N. Strom, S. Matsoukas, and S. Vitaladevuni, “Compressed time delay neural network for small-footprint keyword spotting.” inProc. Interspeech, 2017, pp. 3607–3611

    2017

  9. [17]

    Monophone-based background modeling for two-stage on-device wake word detection,

    M. Wu, S. Panchapagesan, M. Sun, J. Gu, R. Thomas, S. N. P. Vitalade- vuni, B. Hoffmeister, and A. Mandal, “Monophone-based background modeling for two-stage on-device wake word detection,” in Proc. IEEE ICASSP, 2018, pp. 5494–5498

    2018

  10. [18]

    Small-footprint open- vocabulary keyword spotting with quantized LSTM networks,

    T. Bluche, M. Primet, and T. Gisselbrecht, “Small-footprint open- vocabulary keyword spotting with quantized LSTM networks,” CoRR, vol. abs/2002.10851, 2020. [Online]. Available: https://arxiv.org/abs/ 2002.10851

    2002 arXiv

  11. [19]

    Sequence transduction with recurrent neural networks,

    A. Graves, “Sequence transduction with recurrent neural networks,” in ICML — Workshop on Representation Learning , 2012

    2012

  12. [20]

    Intermediate loss regularization for CTC-based speech recognition,

    J. Lee and S. Watanabe, “Intermediate loss regularization for CTC-based speech recognition,” in Proc. IEEE ICASSP , 2021, pp. 6224–6228

    2021

  13. [21]

    Relaxing the conditional independence as- sumption of ctc-based ASR by conditioning on intermediate predictions,

    J. Nozaki and T. Komatsu, “Relaxing the conditional independence as- sumption of ctc-based ASR by conditioning on intermediate predictions,” in Proc. ISCA Interspeech , 2021, pp. 3735–3739

    2021

  14. [22]

    Hierarchical conditional end-to-end ASR with CTC and multi-granular subword units,

    Y . Higuchi, K. Karube, T. Ogawa, and T. Kobayashi, “Hierarchical conditional end-to-end ASR with CTC and multi-granular subword units,” in Proc. IEEE ICASSP , 2022, pp. 7797–7801

    2022

  15. [23]

    Improving massively multilingual ASR with auxiliary CTC objectives,

    W. Chen, B. Yan, J. Shi, Y . Peng, S. Maiti, and S. Watanabe, “Improving massively multilingual ASR with auxiliary CTC objectives,” in Proc. IEEE ICASSP, 2023, pp. 1–5

    2023

  16. [24]

    Multilingual audio-visual speech recognition with hybrid CTC/RNN-T fast conformer,

    M. Burchi, K. C. Puvvada, J. Balam, B. Ginsburg, and R. Timofte, “Multilingual audio-visual speech recognition with hybrid CTC/RNN-T fast conformer,” inProc. IEEE ICASSP. IEEE, 2024, pp. 10 211–10 215

    2024

  17. [25]

    Librispeech: an asr corpus based on public domain audio books,

    V . Panayotov et al., “Librispeech: an asr corpus based on public domain audio books,” in Proc. IEEE ICASSP , 2015, pp. 5206–5210

    2015

  18. [26]

    Efficient keyword spotting using dilated convolutions and gating,

    A. Coucke, M. Chlieh, T. Gisselbrecht, D. Leroy, M. Poumeyrol, and T. Lavril, “Efficient keyword spotting using dilated convolutions and gating,” in Proc. IEEE ICASSP , 2019, pp. 6351–6355

    2019

  19. [27]

    Wham!: Extending speech separation to noisy environments,

    G. Wichern, J. Antognini, M. Flynn, L. R. Zhu, E. McQuinn, D. Crow, E. Manilow, and J. L. Roux, “Wham!: Extending speech separation to noisy environments,” in Proc. ISCA Interspeech , 2019, pp. 1368–1372

    2019

  20. [28]

    Audio augmentation for speech recognition,

    T. Ko, V . Peddinti, D. Povey, and S. Khudanpur, “Audio augmentation for speech recognition,” in Proc. ISCA Interspeech , 2015

    2015

  21. [29]

    Specaugment: A simple data augmentation method for automatic speech recognition,

    D. S. Park, W. Chan, Y . Zhang, C.-C. Chiu, B. Zoph, E. D. Cubuk, and Q. V . Le, “Specaugment: A simple data augmentation method for automatic speech recognition,” in Proc. ISCA Interspeech , 2019

    2019

  22. [30]

    Deep-FSMN for large vocabulary continuous speech recognition,

    S. Zhang et al. , “Deep-FSMN for large vocabulary continuous speech recognition,” in Proc. IEEE ICASSP . IEEE, 2018, pp. 5869–5873

    2018

  23. [31]

    The CMU pronouncing dictionary,

    “The CMU pronouncing dictionary,” http://www.speech.cs.cmu.edu/ cgi-bin/cmudict

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.