Robust Self-Training System for Spoken Query Information Retrieval using Pitch Range Variations

摘要: This paper presents an Automatic User Profile Building and Training (AUPB&T) system using voice pitch variation for speech recognition engines. The problem with current ASR engines is that their vocabularies are usually only suited general usage. Another there no easy means visually challenged users to train the engine improve its performance. Our proposed solution consists of a will accept user's document favorite web pages. These documents then be parsed words added engine's lexicon. Next, it uses those start training session. completed automatically by high quality text-to-speech (TTS) natural voice. In order overcome limited number TTS voices available, we propose integrate during phase AUPB& T, which cover broader range user variability. results our experiments standard show AUPB&T improved rate unknown s speaker.