Text-to-audio generation synthesizes realistic sounds or music given a natural language prompt. Diffusion-based frameworks, including the Tango and the AudioLDM series, represent the state-of-the-art in text-to-audio generation. Despite achieving high audio fidelity, they incur significant inference latency due to the slow diffusion sampling process. MAGNET, a mask-based model operating on discrete tokens, addresses slow inference through iterative mask-based parallel decoding. However, its audio quality still lags behind that of diffusion-based models. In this work, we introduce IMPACT, a text-to-audio generation framework that achieves high performance in audio quality and fidelity while ensuring fast inference. IMPACT utilizes iterative mask-based parallel decoding in a continuous latent space powered by diffusion modeling. This approach eliminates the fidelity constraints of discrete tokens while maintaining competitive inference speed. Results on AudioCaps demonstrate that IMPACT achieves state-of-the-art performance on key metrics including Fréchet Distance (FD) and Fréchet Audio Distance (FAD) while significantly reducing latency compared to prior models. The project website is available at https://audio-impact.github.io/.

Autoregressive next-token prediction with the Transformer decoder has become a de facto standard in large language models (LLMs), achieving remarkable success in Natural Language Processing (NLP) at scale. Extending this paradigm to audio poses unique challenges due to its inherently continuous nature. We research audio generation with a causal language model (LM) without discrete tokens. We leverage token-wise diffusion to model the continuous distribution of the next continuous-valued token. Our approach delivers significant improvements over previous discrete solution, AudioGen, achieving 20% and 40% relative gains on AudioCaps in Frechet Audio Distance (FAD) and Kullback-Leibler (KL) divergence, respectively. Additionally, we propose a novel masked next-token prediction task that incorporates masked prediction into the causal LM framework. On AudioCaps, the innovation yields 41% and 33% relative FAD improvements over AudioGen Base (285M) and AudioGen Large (1B) models, respectively, and is on par with the state-of-the-art (SOTA) diffusion models. Furthermore, we achieve these results with significantly fewer parameters—193M for our Base and 462M for our Large models.

Recent advances in machine learning have demonstrated that multi-modal pre-training can improve automatic speech recognition (ASR) performance compared to randomly initialized models, even when models are fine-tuned on uni-modal tasks. Existing multi-modal pre-training methods for the ASR task have primarily focused on single-stage pre-training where a single unsupervised task is used for pre-training followed by fine-tuning on the downstream task. In this work, we introduce a novel method combining multi-modal and multi-task unsupervised pre-training with a translation-based supervised mid-training approach. We empirically demonstrate that such a multi-stage approach leads to relative word error rate (WER) improvements of up to 38.45% over baselines on both Librispeech and SUPERB. Additionally, we share several important findings for choosing pre-training methods and datasets.

Ninareh Mehrabi, Palash Goyal, Christophe Dupuy, Qian Hu, Shalini Ghosh, Richard Zemel, Kai-Wei Chang, Aram Galstyan, Rahul Gupta. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

While word error rates of automatic speech recognition (ASR) systems have consistently fallen, natural language understanding (NLU) applications built on top of ASR systems still attribute significant numbers of failures to low-quality speech recognition results. Existing assistant systems collect large numbers of these unsuccessful interactions, but these systems usually fail to learn from these interactions, even in an offline fashion. In this work, we introduce CLC: Contrastive Learning for Conversations, a family of methods for contrastive fine-tuning of models in a self-supervised fashion, making use of easily detectable artifacts in unsuccessful conversations with assistants. We demonstrate that our CLC family of approaches can improve the performance of ASR models on OD3, a new public large-scale semi-synthetic meta-dataset of audio task-oriented dialogues, by up to 19.2%. These gains transfer to real-world systems as well, where we show that CLC can help to improve performance by up to 6.7% over baselines.1

In the realm of spoken language understanding (SLU). numerous natural language understanding (NLU) methodologies have been adapted by supplying large language models (LLMs) with transcribed speech instead of conventional written text. In real-world scenarios, prior to input into an LLM. an automated speech recognition (ASR) system generates an output transcript hypothesis, where inherent errors can degrade subsequent SLU tasks. Here we introduce a method that utilizes the ASR system's lattice output instead of relying solely on the top hypothesis, aiming to encapsulate speech ambiguities and enhance SLU outcomes. Our in-context learning experiments, covering spoken question answering and intent classification. underline the LLM's resilience to noisy speech transcripts with the help of word confusion networks from lattices, bridging the SLU performance gap between using the top ASR hypothesis and an oracle upper bound. Additionally, we delve into the LLM's robustness to varying ASR performance conditions and scrutinize the aspects of in-context learning which prove the most influential.

Large Language Models (LLMs) have demonstrated superior abilities in tasks such as chatting, reasoning, and question-answering. However, standard LLMs may ignore crucial paralinguistic information, such as sentiment, emotion, and speaking style, which are essential for achieving natural, human-like spoken conversation, especially when such information is conveyed by acoustic cues. We therefore propose Paralinguistics-enhanced Generative Pretrained Transformer (ParalinGPT), an LLM that utilizes text and speech modalities to better model the linguistic content and paralinguistic attributes of spoken dialogue. The model takes the conversational context of text, speech embeddings, and paralinguistic attributes as input prompts within a serialized multitasking multimodal framework. Specifically, our framework serializes tasks in the order of current paralinguistic attribute prediction, response paralinguistic attribute prediction, and response text generation with autoregressive conditioning. We utilize the Switchboard-1 corpus, including its sentiment labels as the paralinguistic attribute, as our spoken dialogue dataset. Experimental results indicate the proposed serialized multitasking method outperforms typical sequence classification techniques on current and response sentiment classification. Furthermore, leveraging conversational context and speech embeddings significantly improves both response text generation and sentiment prediction. Our proposed framework achieves relative improvements of 6.7%, 12.0%, and 3.5% in current sentiment accuracy, response sentiment accuracy, and response text BLEU score, respectively.

Dialog systems, such as voice assistants, are expected to engage with users in complex, evolving conversations. Unfortunately, traditional automatic speech recognition (ASR) systems deployed in such applications are usually trained to recognize each turn independently and lack the ability to adapt to the conversational context or incorporate user feedback. In this work, we introduce a general framework for ASR in dialog systems that can go beyond learning from single-turn utterances and learn over time how to adapt to both explicit supervision and implicit user feedback present in multi-turn conversations. We accomplish that by leveraging advances in student-teacher learning and context-aware dialog processing, and designing contrastive self-supervision approaches with Ohm, a new online hard-negative mining approach. We show that leveraging our new framework compared to traditional training leads to relative WER reductions of close to 10% in real-world dialog systems, and up to 26% on public synthetic data.

We explore the ability of large language models (LLMs) to act as speech recognition post-processors that perform rescoring and error correction. Our first focus is on instruction prompting to let LLMs perform these task without fine-tuning, for which we evaluate different prompting schemes, both zeroand few-shot in-context learning, and a novel “task activation“ prompting method that combines causal instructions and demonstration to increase its context windows. Next, we show that rescoring only by in-context learning with frozen LLMs achieves results that are competitive with rescoring by domain-tuned LMs, using a pretrained first-pass recognition system and rescoring output on two out-of-domain tasks (ATIS and WSJ). By combining prompting techniques with fine-tuning we achieve error rates below the N-best oracle level, showcasing the generalization power of the LLMs.

We propose a neural language modeling system based on low-rank adaptation (LoRA) for speech recognition output rescoring. Although pretrained language models (LMs) like BERT have shown superior performance in second-pass rescoring, the high computational cost of scaling up the pretraining stage and adapting the pretrained models to specific domains limit their practical use in rescoring. Here we present a method based on low-rank decomposition to train a rescoring BERT model and adapt it to new domains using only a fraction (0.08%) of the pretrained parameters. These inserted matrices are optimized through a discriminative training objective along with a correlation-based regularization loss. The proposed low-rank adaptation RescoreBERT (LoRB) architecture is evaluated on LibriSpeech and internal datasets with decreased training times by factors between 5.4 and 3.6.

Despite improvements to the generalization performance of automated speech recognition (ASR) models, specializing ASR models for downstream tasks remains a challenging task, primarily due to reduced data availability (necessitating increased data collection), and rapidly shifting data distributions (requiring more frequent model fine-tuning). In this work, we investigate the potential of leveraging external knowledge, particularly through off-policy generated text-to-speech key-value stores, to allow for flexible post-training adaptation to new data distributions. In our approach, audio embeddings captured from text-to-speech are used, along with semantic text embeddings, to bias ASR via an approximate k-nearest-neighbor (KNN) based attentive fusion step. Our experiments on LibiriSpeech and in-house voice assistant/search datasets show that the proposed approach can reduce domain adaptation time by up to 1K GPU-hours while providing up to 3% WER improvement compared to a fine-tuning baseline, suggesting a promising approach for adapting production ASR systems in challenging zero and few-shot scenarios.

The development of large datasets for various tasks has driven the success of deep learning models but at the cost of increased label noise, duplication, collection challenges, storage capabilities, and training requirements. In this work, we investigate whether all samples in large datasets contribute equally to better model accuracy. We study statistical and mathematical techniques to reduce redundancies in datasets by directly optimizing data samples for the generalization accuracy of deep learning models. Existing dataset optimization approaches include analytic methods that remove unimportant samples and synthetic methods that generate new datasets to maximize the generalization accuracy. We develop Prune then distill, a combination of analytic and synthetic dataset optimization algorithms, and demonstrate up to 15% relative improvement in generalization accuracy over either approach used independently on standard image and audio classification tasks. Additionally, we demonstrate up to 38% improvement in generalization accuracy of dataset pruning algorithms by maintaining class balance while pruning.

Traditionally, research in automated speech recognition has focused on local-first encoding of audio representations to predict the spoken phonemes in an utterance. Unfortunately, approaches relying on such hyper-local information tend to be vulnerable to both local-level corruption (such as audio-frame drops, or loud noises) and global-level noise (such as environmental noise, or background noise) that has not been seen during training. In this work, we introduce a novel approach that leverages a self-supervised learning technique based on masked language modeling to compute a global, multi-modal encoding of the environment in which the utterance occurs. We then use a new deep-fusion framework to integrate this global context into a traditional ASR method, and demonstrate that the resulting method can outperform baseline methods by up to 7% on Librispeech; gains on internal datasets range from 6% (on larger models) to 45% (on smaller models).

Deep neural networks have largely demonstrated their ability to perform automated speech recognition (ASR) by extracting meaningful features from input audio frames.Such features, however, may consist not only of information about the spoken language content, but also may contain information about unnecessary contexts such as background noise and sounds or speaker identity, accent, or protected attributes.Such information can directly harm generalization performance, by introducing spurious correlations between the spoken words and the context in which such words were spoken.In this work, we introduce an unsupervised, encoder-agnostic method for factoring speech-encoder representations into explicit content-encoding representations and spurious context-encoding representations.By doing so, we demonstrate improved performance on standard ASR benchmarks, as well as improved performance in both real-world and artificially noisy ASR scenarios.

Zhekun Luo, Shalini Ghosh, Devin Guillory, Keizo Kato, Trevor Darrell, Huijuan Xu. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

Classification algorithms in machine learning often assume a flat label space. However, most real world data have dependencies between the labels, which can often be captured by using a hierarchy. Utilizing this relation can help develop a model capable of satisfying the dependencies and improving model accuracy and interpretability. Further, as different levels in the hierarchy correspond to different granularities, penalizing each label equally can be detrimental to model learning. In this paper, we propose a loss function, hierarchical curriculum loss, with two properties: (i) satisfy hierarchical constraints present in the label space, and (ii) provide non-uniform weights to labels based on their levels in the hierarchy, learned implicitly by the training paradigm. We theoretically show that the proposed hierarchical class-based curriculum loss is a tight bound of 0-1 loss among all losses satisfying the hierarchical constraints. We test our loss function on real world image data sets, and show that it significantly outperforms state-of-the-art baselines.

In this talk, we will first discuss multimodal ML for video content analysis. Videos typically have data in multiple modalities like audio, video, and text (captions). Understanding and modeling the interaction between different modalities is key for video analysis tasks like categorization, object detection, activity recognition, etc. However, data modalities are not always correlated -- so, learning when modalities are correlated and using that to guide the influence of one modality on the other is crucial. Another salient feature of videos is the coherence between successive frames due to continuity of video and audio, a property that we refer to as temporal coherence. We show how using non-linear guided cross-modal signals and temporal coherence can improve the performance of multimodal ML models for video analysis tasks like categorization. We also created a hierarchical taxonomy of categories internally. Our experiments on the large-scale YouTube-8M dataset show how our approach significantly outperforms state-of-the-art multimodal ML model for video categorization using our taxonomy, as well as generalizes well to an internal dataset of video segments from actual TV programs. The next part of the talk will briefly discuss our work on explainability of multimodal ML models. We will conclude the talk by outlining other multimodal ML applications like incremental object detection and visual dialog, and discuss potential applications of multimodal ML to various domains.

Videos have data in multiple modalities, e.g., audio, video, text (captions). Understanding and modeling the interaction between different modalities is key for video analysis tasks like categorization, object detection, activity recognition, etc. However, data modalities are not always correlated --- so, learning when modalities are correlated and using that to guide the influence of one modality on the other is crucial. Another salient feature of videos is the coherence between successive frames due to continuity of video and audio, a property that we refer to as temporal coherence. We show how using non-linear guided cross-modal signals and temporal coherence can improve the performance of multi-modal machine learning (ML) models for video analysis tasks like categorization. Our experiments on the large-scale YouTube-8M dataset show how our approach significantly outperforms state-of-the-art multi-modal ML models for video categorization. The model trained on the YouTube-8M dataset also showed good performance on an internal dataset of video segments from actual Samsung TV Plus channels without retraining or fine-tuning, showing the generalization capabilities of our model.

Deep neural networks (DNNs) often suffer from "catastrophic forgetting" during incremental learning (IL) — an abrupt degradation of performance on the original set of classes when the training objective is adapted to a newly added set of classes. Existing IL approaches tend to produce a model that is biased towards either the old classes or new classes, unless with the help of exemplars of the old data. To address this issue, we propose a class-incremental learning paradigm called Deep Model Consolidation (DMC), which works well even when the original training data is not available. The idea is to first train a separate model only for the new classes, and then combine the two individual models trained on data of two distinct set of classes (old classes and new classes) via a novel double distillation training objective. The two existing models are consolidated by exploiting publicly available unlabeled auxiliary data. This overcomes the potential difficulties due to unavailability of original training data. Compared to the state-of-the-art techniques, DMC demonstrates significantly better performance in image classification (CIFAR-100 and CUB-200) and object detection (PASCAL VOC 2007) in the single-headed IL setting.

Continual learning (CL), the problem of lifelong learning where tasks arrive in sequence, has attracted increasing attention in the computer vision community lately. The goal of CL is to learn new tasks while maintaining the performance on the previously learned tasks. There are two major obstacles for CL of deep neural networks: catastrophic forgetting and limited model capacity. Inspired by the recent breakthroughs in automatically learning good neural network architectures, we develop a nonexpansive AutoML framework for CL termed Regularize, Expand and Compress (REC) to solve the above issues. REC is a unified framework with three highlights: 1) a novel regularized weight consolidation (RWC) algorithm to avoid forgetting, where accessing the data seen in the previously learned tasks is not required; 2) an automatic neural architecture search (AutoML) engine to expand the network to increase model capability; 3) smart compression of the expanded model after a new task is learned to improve the model efficiency. The experimental results on four different image recognition datasets demonstrate the superior performance of the proposed REC over other CL algorithms.

An important consideration in software-defined networks (SDNs), is that one SDN application, through a bug or API misuse, can break an entire SDN. While previous works have tried to mitigate such concerns by implementing access control mechanisms (permission models) for an SDN controller, they commonly require serious manual efforts in creating a permission model. Moreover, they do not support flexible permission models, and they are often tightly coupled with a specific SDN controller. To address such limitations, we introduce an automated permission generation and verification system called VOGUE. A distinguishing aspect of VOGUE is that it automatically generates flexible permission models and yet is completely separated from the SDN controller implementation. To demonstrate the feasibility of our approach, we implement a prototype, evaluate its completeness and soundness, and examine its performance. In addition, to show the effectiveness of VOGUE, we demonstrate its use cases and security impact to SDN in the context of popular SDN controllers.

Many vision and language models suffer from poor visual grounding -- often falling back on easy-to-learn language priors rather than basing their decisions on visual concepts in the image. In this work, we propose a generic approach called Human Importance-aware Network Tuning (HINT) that effectively leverages human demonstrations to improve visual grounding. HINT encourages deep networks to be sensitive to the same input regions as humans. Our approach optimizes the alignment between human attention maps and gradient-based network importances -- ensuring that models learn not just to look at but rather rely on visual concepts that humans found relevant for a task when making predictions. We apply HINT to Visual Question Answering and Image Captioning tasks, outperforming top approaches on splits that penalize over-reliance on language priors (VQA-CP and robust captioning) using human attention demonstrations for just 6% of the training data.

The key challenge of generative Visual Dialogue (VD) systems is to respond to human queries with informative answers in natural and contiguous conversation flow. Traditional Maximum Likelihood Estimation-based methods only learn from positive responses but ignore the negative responses, and consequently tend to yield safe or generic responses. To address this issue, we propose a novel training scheme in conjunction with weighted likelihood estimation method. Furthermore, an adaptive multi-modal reasoning module is designed, to accommodate various dialogue scenarios automatically and select relevant information accordingly. The experimental results on the VisDial benchmark demonstrate the superiority of our proposed algorithm over other state-of-the-art approaches, with an improvement of 5.81% on recall@10.

Onion sites on the darkweb operate using the Tor Hidden Service (HS) protocol to shield their locations on the Internet, which (among other features) enables these sites to host malicious and illegal content while being resistant to legal action and seizure. Identifying and monitoring such illicit sites in the darkweb is of high relevance to the Computer Security and Law Enforcement communities. We have developed an automated infrastructure that crawls and indexes content from onion sites into a large-scale data repository, called LIGHTS, with over 100M pages. In this paper we describe Automated Tool for Onion Labeling (ATOL), a novel scalable analysis service developed to conduct a thematic assessment of the content of onion sites in the LIGHTS repository. ATOL has three core components -- (a) a novel keyword discovery mechanism (ATOLKeyword) which extends analyst-provided keywords for different categories by suggesting new descriptive and discriminative keywords that are relevant for the categories; (b) a classification framework (ATOLClassify) that uses the discovered keywords to map onion site content to a set of categories when sufficient labeled data is available; (c) a clustering framework (ATOLCluster) that can leverage information from multiple external heterogeneous knowledge sources, ranging from domain expertise to Bitcoin transaction data, to categorize onion content in the absence of sufficient supervised data. The paper presents empirical results of ATOL on onion datasets derived from the LIGHTS repository, and additionally benchmarks ATOL's algorithms on the publicly available 20 Newsgroups dataset to demonstrate the reproducibility of its results. On the LIGHTS dataset, ATOLClassify gives a 12% performance gain over an analyst-provided baseline, while ATOLCluster gives a 7% improvement over state-of-the-art semi-supervised clustering algorithms. We also discuss how ATOL has been deployed and externally evaluated, as part of the LIGHTS system.

The performance and operational characteristics of the Domain Name System (DNS) protocol are of deep interest to the research and network operations community. In this paper, we present measurement results from a unique dataset containing more than 26 billion DNS query-response pairs collected from more than 600 globally distributed recursive DNS resolvers. We use this dataset to reaffirm findings in published work and notice some significant differences that could be attributed both to the evolving nature of DNS traffic and to our differing perspective. For example, we find that although characteristics of DNS traffic vary greatly across networks, the resolvers within an organization tend to exhibit similar behavior. We further find that more than 50% of DNS queries issued to root servers do not return successful answers, and that the primary cause of lookup failures at root servers is malformed queries with invalid top-level domains (TLDs). Furthermore, we propose a novel approach that detects malicious domain groups using temporal correlation in DNS queries. Our approach requires no comprehensive labeled training set, which can be difficult to build in practice. Instead, it uses a known malicious domain as anchor and identifies the set of previously unknown malicious domains that are related to the anchor domain. Experimental results illustrate the viability of this approach, i.e., we attain a true positive rate of more than 96%, and each malicious anchor domain results in a malware domain group with more than 53 previously unknown malicious domains on average.

We present a methodology for the probabilistic modeling of failure dependencies in large, complex systems using Markov Logic Networks (MLNs), a state-of-the-art probabilistic relational modeling technique in machine learning. We illustrate this modeling methodology on example system architectures, and show how the the Probabilistic Consistency Engine (PCE) tool can create and analyze failure-dependency models. We compare MLN-based analysis with analytical symbolic analysis to validate our approach. The latter method yields bounds on the expected system behaviors for different component-failure probabilities, but it requires closed-form representations and is therefore often an impractical approach for complex system analysis. The MLN-based method facilitates techniques of early design analysis for reliability (e.g., probabilistic sensitivity analysis). We analyze two examples - a portion of the Time-Triggered Ethernet (TTEthernet) communication platform used in space, and an architecture based on Honeywell's Cabin Air Compressor(CAC) - that highlight the value of the MLN-based approach for analyzing failure dependencies in complex cyber-physical systems.

The performance and operational characteristics of the DNS protocol are of deep interest to the research and network operations community. In this paper, we present measurement results from a unique dataset containing more than 26 billion DNS query-response pairs collected from more than 600 globally distributed recursive DNS resolvers. We use this dataset to reaffirm findings in published work and notice some significant differences that could be attributed both to the evolving nature of DNS traffic and to our differing perspective. For example, we find that although characteristics of DNS traffic vary greatly across networks, the resolvers within an organization tend to exhibit similar behavior. We further find that more than 50% of DNS queries issued to root servers do not return successful answers, and that the primary cause of lookup failures at root servers is malformed queries with invalid TLDs. Furthermore, we propose a novel approach that detects malicious domain groups using temporal correlation in DNS queries. Our approach requires no comprehensive labeled training set, which can be difficult to build in practice. Instead, it uses a known malicious domain as anchor, and identifies the set of previously unknown malicious domains that are related to the anchor domain. Experimental results illustrate the viability of this approach, i.e. , we attain a true positive rate of more than 96%, and each malicious anchor domain results in a malware domain group with more than 53 previously unknown malicious domains on average.

Markov logic is a rich language that allows one to specify a knowledge base as a set of weighted first-order logic formulas, and to define a probability distribution over truth assignments to ground atoms using this knowledge base. Usually, the weight of a formula cannot be related to the probability of the formula without taking into account the weights of the other formulas. In general, this is not an issue, since the weights are learned from training data. However, in many domains ( e .g. healthcare, dependable systems, e tc.), only little or no training data may be available, but one has access to a domain expert whose knowledge is available in the form of subjective probabilities. Within the framework of Bayesian statistics, we present a formalism for using a domain expert’s knowledge for weight learning. Our approach defines priors that are different from and more general than previously used Gaussian priors over weights. We show how one can learn weights in an MLN by combining subjective probabilities and training data, without requiring that the domain expert provides consistent knowledge. Additionally, we also provide a formalism for capturing conditional subjective probabilities, which are often easier to obtain and more reliable than non-conditional probabilities. We demonstrate the effectiveness of our approach by extensive experiments in a domain that models failure dependencies in a cyber-physical system. Moreover, we demonstrate the advantages of using our proposed prior over that of using non-zero mean Gaussian priors in a commonly cited social network MLN testbed. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

An automated design procedure is described for synthesizing circuits with low power concurrent error detection. It is based on pre-synthesis selection of a parity-check code followed by structure constrained logic optimization that produces a circuit in which all single point faults are guaranteed to be detected. Two new contributions over previous work include (1) the use of a k-way partitioning algorithm combined with local search to select a parity-check code, and (2) a methodology for minimizing power consumption in the CED circuitry. Results indicate significant reductions in area overhead due to the new code selection procedure as well as the ability to find low power implementations for use in power conscious applications.

An interconnect break is a break that occurs in the interconnect wiring, which results in logic gate inputs being disconnected from the drivers and causes the wire to float. Interconnect breaks are the most common types of breaks in modern $CMOS$ integrated circuits, so testing and detecting these breaks has become very important. This paper proposes a model by which standard tests for stuck-at-faults can be used to detect interconnect breaks in a circuit. We do a worst-case analysis of the detection of these breaks and calculate the minimum number of test vectors required to detect breaks with a specified confidence level, using n-detection principles. To enhance the understanding of the breaks in the circuit, we present a statistical model based on the length distribution of the wires surrounding the floating wire where the break occurs. From the model we compute the detection probabilities of such breaks and show that the worst case of detection is when the bias voltage is the same as the logic threshold voltage.

In this paper, a technique for weighted pseudo-random built-in self-test (BIST) of VLSI circuits is proposed, which uses special scan cells and a new weight selection algorithm to achieve low power dissipation. It is based on weighted pseudo-random scan testing in which only 3 weight values are used - 2 fixed values (0 or 1) and 1 random value (0.5). A new weight selection algorithm is used to select a set of weights that achieves high fault coverage while reducing power. The idea is to minimize power by careful selection of the set of scan cells having fixed values (0 or 1) in order to reduce switching activity. To implement this in hardware, a new scan cell design is proposed that can do scan and capture in the normal mode as well as fixed-bit mode. The new scan cell hardware increases the area of a typical circuit by less than 4%, but reduces power by as much as 96%, as indicated in experiments performed on benchmark circuits.

A method is proposed for reducing power consumption in memory ECC checker circuitry that provides SEC-DED. The degrees of freedom in selecting the parity check matrix are used to minimize power with little or no impact on area and delay. The power minimization method is applied to two popular SEC-DED codes: standard Hamming codes and odd-column-weight Hsiao codes. Experiments on actual memory traces of Spec and MediaBench benchmarks indicate that considering power in addition to area and delay when selecting the parity check matrix can result in power reductions of up to 27% for Hsiao codes and up to 41% for Hamming codes.