Speech Recognition AI v20220719

Releasing the ninth version of "Speech Recognition AI".

New Added Features:

• Fixed No Hardware Microphone Issue (uses IP Mic only in that case)

It's just a 127kB JAR file.

Speech Recognition AI v20220707

Releasing the eighth version of "Speech Recognition AI".

New Added Features:

• Server socket timeouts in case of connection failure

It's just a 127kB JAR file.

Speech Recognition AI v20220704

Releasing the seventh version of "Speech Recognition AI".

New Added Features:

• Text Server

It's just a 127kB JAR file.

Speech Recognition AI v20220601

Releasing the sixth version of "Speech Recognition AI".

New Added Features:

• Word Commands: Lets you easily translate a phrase or a word into a different command to control apps and IoT devices.
• Word Responses: Lets you to respond a meaningful reply to a spoken query.

Word Commands

Word Responses

Settings

It is just a 124kB JAR file.

Speech Recognition AI v20220409

Releasing the fifth version of "Speech Recognition AI".

New Added Features:

• Parallel Feed Forward (5 times faster) and Back Propagation (over 9 times faster - 6 core intel i7)
• Adaptive learning rate and momentum (updates based on average loss)
• Learning Loss Chart of Current Loss and Average Loss (displays every 16th sample)

Vocabulary

New Training (learning from scratch)

Transfer Learning (learning continuation)

It is just a 107kB JAR file.

Speech Recognition AI v20220327

Releasing the forth version of "Speech Recognition AI".

New Added Features:

• Remote Android App "IP Mic" IP Mic
• Increased the limit of hidden layers to 10 (12 layers max in total)
• Audio data chart more efficient
• New Advanced Settings options for IP Mic
• Remote IP Mic protected with a token
• Fixed Advanced Setting Layout (on Linux and also Windows)

It is just a 106kB JAR file.

Speech Recognition AI v20220126

Releasing the third version of "Speech Recognition AI".

New Added Features:

• Training Autosaves Parameters (weights and biases) when it lowers loss by the factor of 2 (saves at 0.5, 0.25, 0.125, etc.)
• Exits training and saves final weights when the minimal "Average Loss" is reached
• Advanced Transfer Learning (even if output neurons or last layers do not match)
• Creates new neurons with random parameter values (random weights and biases)
• Renamed "Error" to "Loss"
• shortened the time to stop a recording to 2 seconds (when a user stops talking)

It is just a 100kB JAR file.

Speech Recognition AI v20210927

Releasing the second version of "Speech Recognition AI".
New Added Features:

• Word Detection settings
• Neural Network Visualization option
• Keep Long Word option
• Neural Network Training settings
• Auto-Creating Missing Folder and Files
• Simple Transfer Learning (Only if topology perfectly matches)

It is just a 100kB JAR file.

Speech Recognition AI v20210403

Releasing the first version of "Speech Recognition AI".
Features:

• 16bit mono 22kHz audio sampling
• automatically starts to record and stops recording
• 22 second buffer for speech and 2.97 second buffer word (phrase)
• word routing: Allows to send recognized speech as text to another applications (IoT, speech control)