Making ideal decisions as a product leader in a web-facing company is incredibly challenging. Beyond navigating the ambiguity of customer satisfaction and achieving business goals, leaders must also ensure their products and services remain relevant, desirable, and profitable. Data and experimentation are crucial for testing product hypotheses and informing decisions. Online controlled experiments, such as A/B testing, can provide highly reliable data to support decisions. However, these experiments can be time-consuming and costly, particularly when assessing impacts on key business metrics like retention or long-term value.

This paper describes the Microsoft Text-to-Speech (TTS) system: LeanSpeech for LIMMITS (Lightweight, Multi-speaker, Multi-lingual Indic TTS) Challenge 20231, which is part of ICASSP2023 to encourage the advance of TTS in Indian Languages. We propose a lightweight encoder-decoder acoustic model composed of 1-D convolution and LSTM blocks, which is trained with knowledge distillation from a multi-speaker multi-lingual teacher model, DelightfulTTS [1]. The speech corpus is reprocessed and used in both AM training and vocoder fine-tuning. In Track-2 of the challenge, our system achieves MOS 4.56 and SMOS 3.98, which indicates the efficiency of the proposed model and training strategy.

American Sign Language (ASL), an essential communication tool for the deaf and hard of hearing, has long relied on traditional teaching methods that may not fully leverage the potential of modern technology. Recognizing this gap, this paper introduces VirtuSign, an innovative solution that merges virtual reality (VR) with advanced gesture recognition to revolutionize ASL instruction. VirtuSign offers a tailored VR experience, immersing users in a gamified environment where they can interactively learn and practice signing the ASL alphabet. Enhanced by state-of-the-art artificial intelligence algorithms, the tool identifies and assesses user gestures in real-time through hand-tracking capabilities available in modern VR headsets, providing immediate feedback and corrections.

The end-to-end (E2E) approach to automatic speech recognition (ASR) is a simplified and an elegant approach where a single deep neural network model directly converts the acoustic feature sequence to the text sequence. The current approach to end-to-end ASR uses the neural network model (trained with sequence loss) along with an external character/word based language model (LM) in a decoding pass to output the text sequence. In this work, we propose a new objective function for end-to-end ASR training where the LM score is explicitly introduced in the attention model loss function without any additional training parameters. In this manner, the neural network is made LM aware and this simplifies the model training process. We also propose to incorporate an attention based sequence summary feature in the ASR model which allows the system to be speaker aware. With several E2E ASR experiments on TED-LIUM, WSJ and Librispeech datasets, we show that the proposed speaker and LM aware training improves the ASR performance significantly over the state-of-art E2E approaches. We achieve the best published results reported for WSJ dataset.

Recently developed end-to-end (E2E) automatic speech recognition (ASR) systems demand abundance of transcribed speech data, there are several scenarios where the labeling of speech data is cumbersome and expensive. For a fixed annotation cost, active learning for speech recognition allows to efficiently train the ASR model. In this work, we advance the most common approach for active learning methods which relies on uncertainty sampling technique. In particular, we explore the use of path probability of the decoded sequence as a confidence measure and select the samples with the least confidence for active learning. In order to reduce the sampling bias in active learning, we propose a regularized uncertainty sampling approach that incorporates an i-vector diversity measure. Thus, the active learning in the proposed framework uses a joint score of uncertainty and i-vector diversity. The benefits of the proposed approach are illustrated for an E2E ASR task performed on CSJ and Librispeech datasets. In these experiments, we show that the proposed approach yields considerable improvements over the baseline model using random sampling.

In this paper, we propose a novel approach to a challenging problem of daily life cooking activity recognition task based upon object use and frame sequence tagging. We use a dynamic SVM-HMM hybrid model which combines structural as well as temporal video sequence information to jointly infer the most likely cooking activity labels. We demonstrate that our approach can achieve activity recognition rates for kitchen scenarios of more than 72% on a real-world cooking dataset consisting of 9 cooking activities with significant variations in performance of these activities by different subjects. Such a context based approach as discussed in this paper can be extended to other fine grain activities such as hospital operating rooms in medical practices, agricultural and manufacturing operations, etc.

Human Computer Interaction is an upcoming scientific field which aims at inter-communication between humans and computers. A major element of this field is Human Emotion Recognition. The most expressive way humans display emotions is through facial expressions. Traditionally, emotion recognition has been performed on laboratory controlled data. While undoubtedly worthwhile at the time, such lab controlled data poorly represents the environment and conditions faced in real-world situations. With the increase in the number of video clips online, it is worthwhile to explore the performance of emotion recognition methods that work 'in the wild' .This work mainly focuses on automatic emotion recognition in a wild video sample. In this task, we have worked on the problem of human emotion recognition using a combination of video features and audio features. The technique that we have utilized for emotion detection involves a blend of Optical flow, Gabor Filtering, few other facial features and audio features. Training and Classification is performed using Support Vector Machine-Hidden Markov Model (HMM). The unique thing about our methodology is that it produces better results for some particular class of emotions as compared to the baseline score in the case of wild emotion dataset with an overall accuracy of 20.51% on the test set.