Has anyone tried Snowboy with telephony speech (instead of using mic input) using any of the VoIP codecs like AMR-WB, EVS or OPUS? We are trying feeding decoded speech buffer to Snowboy and seeing ~50% drop in detection ratio v/s when using mic input directly (tried with universal model "alexa"). There may be some issue with our application code, trying to figure it out. Please share experience if someone has already tried it.
Has anyone tried Snowboy with telephony speech (instead of using mic input) using any of the VoIP codecs like AMR-WB, EVS or OPUS? We are trying feeding decoded speech buffer to Snowboy and seeing ~50% drop in detection ratio v/s when using mic input directly (tried with universal model "alexa"). There may be some issue with our application code, trying to figure it out. Please share experience if someone has already tried it.