Voice Authentication App using Microsoft Cognitive Services

An app that logs users in by voice. Users enrol a passphrase with Microsoft's Speaker Verification API and later verify by speaking it, and can then manage their notes.

The user installs the app and signs in; credentials are checked against Firebase Authentication or the app's own API. Screens built in XML (or Jetpack Compose) collect the user's input and show results in lists and cards. Data is stored in Firebase Realtime Database / Firestore or a local SQLite database and synced when the phone is online. Notifications and background services alert the user to updates, and an admin side manages the records.

An Activity moves through onCreate, onStart, onResume (running), onPause, onStop and onDestroy. Saving state in onPause or onSaveInstanceState prevents data loss when the user switches apps or rotates the screen.

Firebase gives a hosted real-time database, authentication and push notifications without running our own server. SQLite stores data locally on the phone and works offline. We chose based on whether data must be shared between users.

Lowercasing, removing punctuation and stop words, tokenisation and stemming or lemmatisation. The cleaned text is then converted to numbers with bag-of-words, TF-IDF or embeddings.

Term frequency–inverse document frequency gives a word a high weight if it appears often in one document but rarely across all documents, so common words like 'the' get low importance.

Symmetric encryption (AES, DES) uses one shared key for encryption and decryption and is fast. Asymmetric encryption (RSA) uses a public key to encrypt and a private key to decrypt, which solves key sharing but is slower.

We followed an iterative (incremental) model: requirements and a basic design first, then modules built and tested one by one, with the guide's feedback after each review. It suits a project where requirements become clearer as we build.

Unit testing of each module, integration testing of the complete flow and system testing with real inputs. Test cases with expected and actual results are listed in the report, including invalid inputs.

The current version is tested on a limited dataset or number of users. Future work includes deploying on the cloud, adding more features or languages, improving accuracy with more data and building a mobile version.