Skip to main content

Arousal Model

The Arousal model can classify speech into "low", "neutral" or "high". This can be interpreted as the energy in the voice. Tired and depressed speakers would speak with low arousal, whereas angry or excited speakers would present with high arousal in their voice.

Because it only makes sense to apply this model to speech audio, it is combined with the SpeechRT and Volume models to increase the reliability of the results.

The receptive field of this model is 2107 milliseconds.


Receptive FieldResult Type
2107 msresult ∈ ["high", "low", "neutral", "no_speech", "silence"]


The time-series result will be an iterable with elements that contain the following information:

{  "timestamp": 0,  "results":{    "arousal": {        "result": "high",          "confidence": 0.92    }  }}

Time-series with raw values#

If raw values were requested, they will be added to the time-series result:

{  "timestamp": 0,  "results":{    "arousal": {        "result": "high",          "confidence": 0.92    }  },  "raw": {    "arousal": {      "high": 0.92,      "low": 0.0073,      "neutral": 0.0727    }  }}


In case a summary is requested the following will be returned

{  "arousal": {    "high_fraction": 0.30,    "low_fraction": 0.60,    "neutral_fraction": 0.05,    "no_speech_fraction": 0.05,    "silence_fraction": 0.0  }}

where x_fraction represents the percentage of time that x class was identified for the duration of the input.


In case the transitions are requested a time-series with transition elements like shown below will be returned

{  "timestamp_start": 0,  "timestamp_end": 1500,  "result": "neutral",  "confidence": 0.96},{  "timestamp_start": 1500,  "timestamp_end": 4000,  "result": "high",  "confidence": 0.88}

The example above means that the first 1500ms of the audio snippet contained neutral speech, and between 1500ms and 4000ms DeepTone™ detected high arousal in the voice.