Gaussian processes; acoustic signal processing; feedforward neural nets; hidden Markov models; learning (artificial intelligence); natural language processing; speaker recognition; speech processing; DNN training; GMM; Gaussian mixture models; HMM; Hindi LVCSR; Hindi speech recognition; Hindi-English phone pairs; Indian languages; Wall Street Journal English data; acoustic modeling; baseline Gaussian model-sharing approach; data sharing; deep neural network; feed-forward network; hidden Markov models; low-resource settings; phone recognition tasks; phonetic mapping; Acoustics; Data models; Feature extraction; Hidden Markov models; Speech; Speech recognition; Training; Hindi LVSCR; data borrowing; low resource; phone mapping;
机译:集成多种声学和语言模型以改进印地语语音识别系统
机译:在词典中纳入更精细的声学语音特征以进行印地语语音识别
机译:使用内插复发性神经网络语言建模判别训练的连续印地语语音识别系统
机译:低资源设置中印地语语音识别的声学建模
机译:用于低资源和形态复杂语言的自动语音识别
机译:用BPE-ropout进行动态声学单元增强用于低资源端到端语音识别
机译:基于深度学习的低资源语音识别的声学建模:概述