Peter Ochieng

Peter Ochieng

Deep neural networks (DNN) techniques have become pervasive in domains such as natural language processing and computer vision. They have achieved great success in these domains in task such as machine translation and image generation. Due to their success, these data driven techniques have been applied in audio domain. More specifically, DNN models have been applied in speech enhancement domain to achieve denosing, dereverberation and multi-speaker separation in monaural speech enhancement. In this paper, we review some dominant DNN techniques being employed to achieve speech separation. The review looks at the whole pipeline of speech enhancement from feature extraction, how DNN based tools are modelling both global and local features of speech and model training (supervised and unsupervised). We also review the use of speech-enhancement pre-trained models to boost speech enhancement process. The review is geared towards covering the dominant trends with regards to DNN application in speech enhancement in speech obtained via a single speaker.

Peter Ochieng, Dennis Kaburu

Machine hearing of the environmental sound is one of the important issues in the audio recognition domain. It gives the machine the ability to discriminate between the different input sounds that guides its decision making. In this work we exploit the self-supervised contrastive technique and a shallow 1D CNN to extract the distinctive audio features (audio representations) without using any explicit annotations.We generate representations of a given audio using both its raw audio waveform and spectrogram and evaluate if the proposed learner is agnostic to the type of audio input. We further use canonical correlation analysis (CCA) to fuse representations from the two types of input of a given audio and demonstrate that the fused global feature results in robust representation of the audio signal as compared to the individual representations. The evaluation of the proposed technique is done on both ESC-50 and UrbanSound8K. The results show that the proposed technique is able to extract most features of the environmental audio and gives an improvement of 12.8% and 0.9% on the ESC-50 and UrbanSound8K datasets respectively.

Peter Ochieng

This work introduces UDPNet, a novel architecture designed to accelerate the reverse diffusion process in speech synthesis. Unlike traditional diffusion models that rely on timestep embeddings and shared network parameters, UDPNet unrolls the reverse diffusion process directly into the network architecture, with successive layers corresponding to equally spaced steps in the diffusion schedule. Each layer progressively refines the noisy input, culminating in a high-fidelity estimation of the original data, \(x_0\). Additionally, we redefine the learning target by predicting latent variables instead of the conventional \(x_0\) or noise \(ε_0\). This shift addresses the common issue of large prediction errors in early denoising stages, effectively reducing speech distortion. Extensive evaluations on single- and multi-speaker datasets demonstrate that UDPNet consistently outperforms state-of-the-art methods in both quality and efficiency, while generalizing effectively to unseen speech. These results position UDPNet as a robust solution for real-time speech synthesis applications. Sample audio is available at https://onexpeters.github.io/UDPNet.

Micha Kaiser, Paul Lohmann, Peter Ochieng et al.

Food consumption and production contribute significantly to global greenhouse gas emissions, making them crucial entry points for mitigating climate change and maintaining a liveable planet. Over the past two decades, food policy initiatives have explored interventions to reshape production and consumption patterns, focusing on reducing food waste and curbing ruminant meat consumption. While the evidence of "what works" improves, evaluating which policies are appropriate and effective in specific contexts remains difficult due to external validity challenges. This paper demonstrates that a fine-tuned large language model (LLM) can accurately predict the direction of outcomes in approximately 80\% of empirical studies measuring dietary-based impacts (e.g. food choices, sales, waste) resulting from behavioral interventions and policies. Approximately 75 prompts were required to achieve optimal results, with performance showing signs of catastrophic loss beyond this point. Our findings indicate that greater input detail enhances predictive accuracy, although the model still faces challenges with unseen studies, underscoring the importance of a representative training sample. As LLMs continue to improve and diversify, they hold promise for advancing data-driven, evidence-based policymaking.

Peter Ochieng

We derive non-asymptotic spectral bands that bound the squared InfoNCE gradient norm via alignment, temperature, and batch spectrum, recovering the \(1/τ^{2}\) law and closely tracking batch-mean gradients on synthetic data and ImageNet. Using effective rank \(R_{\mathrm{eff}}\) as an anisotropy proxy, we design spectrum-aware batch selection, including a fast greedy builder. On ImageNet-100, Greedy-64 cuts time-to-67.5\% top-1 by 15\% vs.\ random (24\% vs.\ Pool--P3) at equal accuracy; CIFAR-10 shows similar gains. In-batch whitening promotes isotropy and reduces 50-step gradient variance by \(1.37\times\), matching our theoretical upper bound.

Peter Ochieng, Dennis Kaburu

We present a prosody-guided framework for speech-to-speech translation (S2ST) that aligns and translates speech \emph{without} transcripts by leveraging cross-linguistic pause synchrony. Analyzing a 6{,}000-hour East African news corpus spanning five languages, we show that \emph{within-phylum} language pairs exhibit 30--40\% lower pause variance and over 3$\times$ higher onset/offset correlation compared to cross-phylum pairs. These findings motivate \textbf{SPaDA}, a dynamic-programming alignment algorithm that integrates silence consistency, rate synchrony, and semantic similarity. SPaDA improves alignment $F_1$ by +3--4 points and eliminates up to 38\% of spurious matches relative to greedy VAD baselines. Using SPaDA-aligned segments, we train \textbf{SegUniDiff}, a diffusion-based S2ST model guided by \emph{external gradients} from frozen semantic and speaker encoders. SegUniDiff matches an enhanced cascade in BLEU (30.3 on CVSS-C vs.\ 28.9 for UnitY), reduces speaker error rate (EER) from 12.5\% to 5.3\%, and runs at an RTF of 1.02. To support evaluation in low-resource settings, we also release a three-tier, transcript-free BLEU suite (M1--M3) that correlates strongly with human judgments. Together, our results show that prosodic cues in multilingual speech provide a reliable scaffold for scalable, non-autoregressive S2ST.

Peter Ochieng

The current monaural state of the art tools for speech separation relies on supervised learning. This means that they must deal with permutation problem, they are impacted by the mismatch on the number of speakers used in training and inference. Moreover, their performance heavily relies on the presence of high-quality labelled data. These problems can be effectively addressed by employing a fully unsupervised technique for speech separation. In this paper, we use contrastive learning to establish the representations of frames then use the learned representations in the downstream deep modularization task. Concretely, we demonstrate experimentally that in speech separation, different frames of a speaker can be viewed as augmentations of a given hidden standard frame of that speaker. The frames of a speaker contain enough prosodic information overlap which is key in speech separation. Based on this, we implement a self-supervised learning to learn to minimize the distance between frames belonging to a given speaker. The learned representations are used in a downstream deep modularization task to cluster frames based on speaker identity. Evaluation of the developed technique on WSJ0-2mix and WSJ0-3mix shows that the technique attains SI-SNRi and SDRi of 20.8 and 21.0 respectively in WSJ0-2mix. In WSJ0-3mix, it attains SI-SNRi and SDRi of 20.7 and 20.7 respectively in WSJ0-2mix. Its greatest strength being that as the number of speakers increase, its performance does not degrade significantly.

Peter Ochieng

This paper looks at the ability of large language models to participate in educational guided reading. We specifically, evaluate their ability to generate meaningful questions from the input text, generate diverse questions both in terms of content coverage and difficulty of the questions and evaluate their ability to recommend part of the text that a student should re-read based on the student's responses to the questions. Based on our evaluation of ChatGPT and Bard, we report that, 1) Large language models are able to generate high quality meaningful questions that have high correlation with the input text, 2) They generate diverse question that cover most topics in the input text even though this ability is significantly degraded as the input text increases, 3)The large language models are able to generate both low and high cognitive questions even though they are significantly biased toward low cognitive question, 4) They are able to effectively summarize responses and extract a portion of text that should be re-read.

Peter Ochieng, Dennis Mugambi

Conversational machine reading (CMR) tools have seen a rapid progress in the recent past. The current existing tools rely on the supervised learning technique which require labeled dataset for their training. The supervised technique necessitates that for every new rule text, a manually labeled dataset must be created. This is tedious and error prone. This paper introduces and demonstrates how unsupervised learning technique can be applied in the development of CMR. Specifically, we demonstrate how unsupervised learning can be used in rule extraction and entailment modules of CMR. Compared to the current best CMR tool, our developed framework reports 3.3% improvement in micro averaged accuracy and 1.4 % improvement in macro averaged accuracy.