Sökning: "stora talare"

Visar resultat 1 - 5 av 9 uppsatser innehållade orden stora talare.

  1. 1. Repairing Swedish Automatic Speech Recognition

    Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

    Författare :Karla Rehn; [2021]
    Nyckelord :Automatic speech recognition; Dialogue systems; Language models; ASR Repair; Automatisk taligenkänning; Dialogsystem; Språkmodeller; Reparation av taligenkänning;

    Sammanfattning : The quality of automatic speech recognition has increased dramatically the last few years, but the performance for low and middle resource languages such as Swedish is still far from optimal. In this project a language model trained on large written corpora called KB-BERT is utilized to improve the quality of transcriptions for Swedish. LÄS MER

  3. 2. Experiments in speaker diarization using speaker vectors

    Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

    Författare :Ming Cui; [2021]
    Nyckelord :Speaker Diarization; Embedding Extraction Module; Deep Learning; Supervised method; Unsupervised method; Talardiarisering; inbäddning av extraktionsmodul; djupinlärning; övervakad metod; oövervakad metod;

    Sammanfattning : Speaker Diarization is the task of determining ‘who spoke when?’ in an audio or video recording that contains an unknown amount of speech and also an unknown number of speakers. It has emerged as an increasingly important and dedicated domain of speech research. LÄS MER

  4. 3. Multilayered Lamination for Acoustic Board Manufacturing

    Master-uppsats, Lunds universitet/Innovation

    Författare :Angelica Madeland; [2019]
    Nyckelord :Acoustic panels; lamination techniques; sound absorption; insulation; Technology and Engineering;

    Sammanfattning : Saint-Gobain Ecophon manufactures acoustic ceiling tiles made primarily of glass wool and are interested in exploring the idea of a new manufacturing technique – using thinner layers of material and laminating these together. The possible benefits are, amongst others, the ability to use different kinds of material where they are the most useful, reduced costs, larger amounts of recycled material, less material waste, and more customizability. LÄS MER

  5. 4. Generating Training Data for Keyword Spotting given Few Samples

    Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

    Författare :Pius Friesch; [2019]
    Nyckelord :;

    Sammanfattning : Speech recognition systems generally need a large quantity of highly variable voice and recording conditions in order to produce robust results. In the specific case of keyword spotting, where only short commands are recognized instead of large vocabularies, the resource-intensive task of data acquisition has to be repeated for each keyword individually. LÄS MER

  7. 5. Demonstrative contrasts in Hindukush Indo-Aryan

    Magister-uppsats, Stockholms universitet/Institutionen för lingvistik

    Författare :Noa Lange; [2016]
    Nyckelord :Demonstrative; Hindukush Indo-Aryan; invisible; areal typology; The Greater Hindukush; Demonstrativor; Hindukush-indoariska; osynlig; arealtypologi; The Greater Hindukush;

    Sammanfattning : Hindukush Indo-Aryan (HKIA) is a disputed subgroup of Indo-Aryan languages spoken within a linguistically diverse area stretching from northeastern Afghanistan, across northern Pakistan to northwestern India, principally covering the mountainous region of Hindukush–Karakoram–Western Himalaya. A noteworthy feature of some of these languages is a three-way demonstrative system, or three deictic terms used by speakers to direct one another’s attention to referents at different distances in their environment. LÄS MER