Voice profiles for Speakeasy

Voice profiles improve the accuracy of speech recognition by recording the way that a user says a term. These recordings help Speakeasy match what a user says to the terms that it is expecting.

To use voice profiles, the device must have version 2.1.35 of the Velocity Console and the Android Velocity Client installed. The device must also have version 1.0.26 of Speakeasy.

To use voice profiles with Speakeasy:

1.In the Velocity Console, the administrator enables the Voice profiles option for the host profile and provide the terms that you want the user to train on.

2.The administrator deploys the project and distributes the deployment files (including the Speakeasy files) to the device.

3.When a user is connected to the host, the Voice Training option is available in the Client menu. The user records the terms that the engine doesn't recognize well and saves the profile with a name. The profile is automatically applied.

The voice profile is disabled when all sessions are disconnected, but subsequent sessions can enable the voice profile, or erase the voice profile and create a new one. You can only have one voice profile on the device at a time.

Using voice profiles in Velocity does NOT require the !slot command in the grammar file. As long as the grammar files have the terms exactly as you want to train them, you do not need to modify the grammar files to use voice profiles in Velocity.

If there are multiple host profiles on the device with different training lists, the list for the currently active host profile is what the user trains on.

1.Open a host profile in the Velocity Console.

2.On the Voice tab, make sure Speakeasy is enabled.

3.Enable the Voice profiles option.

4.Use the Add button to add or remove terms that the user can train on. The terms must exactly match terms in a grammar file on the device. The user does not need to train on every term, but they will have to skip any term they don’t want to train on, so try to keep the list as short as possible. The possible terms may include:

•single-digit numbers (0-9).

•words that are in a grammar file.

•phrases that are in quotation marks in a grammar file.

Each term specified for training must match a term in a grammar file exactly, including case. If the word or phrase you provide for training does not match a grammar file, it will not return a result. We suggest opening the grammar files and copying the terms out of them to make sure they match.

After you have configured the host profile to enable voice profiles and provided the terms a user can train on, deploy the project and distribute it to devices.

After voice profiles are enabled, the Voice profile option appears in the menu after the session is connected. Voice training is not mandatory. The voice training goes through the entire list of terms that the administrator provided, but the user is allowed to skip terms and only train on the terms that are problematic.

1.Connect to a host.

2.Tap Menu > Voice profile.

3.Tap Record and say the term on the screen. You can record each term up to 3 times.

If you do not want to record term or you do not want to record a term more than once, tap Skip

4.When you’ve progressed through all the training terms, you are prompted to name the profile. Provide a name and tap Save and apply.

The voice training page closes and you are returned to the active session. The voice profile is applied immediately for any sessions that have the voice profile feature enabled. The voice profile is disabled when all sessions are disconnected.

1.Connect to a host.

2.Tap Menu > Voice profile.

•To disable the existing profile, select None from the list and tap Apply.

•To use the existing profile, select the option with the name of the profile and tap Apply.

•To replace the existing voice profile, select Create new voice profile to replace existing and tap Apply.